Post on 04-May-2019
NBER WORKING PAPER SERIES
URBAN INEQUALITY
Edward L. GlaeserMatthew G. Resseger
Kristina Tobio
Working Paper 14419http://www.nber.org/papers/w14419
NATIONAL BUREAU OF ECONOMIC RESEARCH1050 Massachusetts Avenue
Cambridge, MA 02138October 2008
All three authors thank the Taubman Center for State and Local Government for financial support. The views expressed herein are those of the author(s) and do not necessarily reflect the views of theNational Bureau of Economic Research.
NBER working papers are circulated for discussion and comment purposes. They have not been peer-reviewed or been subject to the review by the NBER Board of Directors that accompanies officialNBER publications.
© 2008 by Edward L. Glaeser, Matthew G. Resseger, and Kristina Tobio. All rights reserved. Shortsections of text, not to exceed two paragraphs, may be quoted without explicit permission providedthat full credit, including © notice, is given to the source.
Urban InequalityEdward L. Glaeser, Matthew G. Resseger, and Kristina TobioNBER Working Paper No. 14419October 2008JEL No. H0,I0,J0,R0
ABSTRACT
What impact does inequality have on metropolitan areas? Crime rates are higher in places with moreinequality, and people in unequal cities are more likely to say that they are unhappy. There is alsoa negative association between local inequality and the growth of both income and population, oncewe control for the initial distribution of skills. What determines the degree of inequality across metropolitanareas? Twenty years ago, metropolitan inequality was strongly associated with poverty, but today,inequality is more strongly linked to the presence of the wealthy. Inequality in skills can explain aboutone third of the variation in income inequality, and that skill inequality is itself explained by historicalschooling patterns and immigration. There are also substantial differences in the returns to skill, relatedto local concentrations in different industries, and these too are strongly correlated with inequality.
Edward L. GlaeserDepartment of Economics315A Littauer CenterHarvard UniversityCambridge, MA 02138and NBEReglaeser@harvard.edu
Matthew G. RessegerENTER POSTAL ADDRESS HEREresseger@fas.harvard.edu
Kristina TobioKennedy School of Government79 JFK St- T347Cambridge, MA 02138kristina_tobio@ksg.harvard.edu
2
I. Introduction
For much of the almost 2,500 years since Plato wrote that “any city however small, is in
fact divided into two, one the city of the poor, the other of the rich,” urban scholars have
been struck by the remarkable amount of income inequality within dense cities (Wheeler,
2005). While there is certainly plenty of rural inequality as well, the density of cities and
urban regions makes the contrast of rich and poor particularly striking. Figure 1 shows
the 45 percent correlation between density and income inequality, measured with the Gini
coefficient, across counties with more than one person per every two acre. The tendency
of dense places to be more unequal motivates this survey of inequality in metropolitan
areas, multi-county units containing a dense agglomeration of population.
America is, on the whole, relatively unequal for a developed country (Alesina and
Glaeser, 2004), but there are some places within the U.S. that are a lot more equal than
others. While Manhattan is the physical embodiment of big-city inequality and has a
Gini coefficient of .6, the Gini coefficient of Kendall County, Illinois is only about one-
half that amount. Kendall is a small but rapidly growing county on the outskirts of the
Chicago area that combines agriculture with a growing presence of middle-income
suburbanites. In Kendall, 9.2 percent of households earned more than 150,000 dollars in
2006, and 9.3 percent of households earned less than 25,000 dollars. In contrast, 20.4
percent of households in New York County earned more than 150,000 dollars, and 26.5
percent earned less than 25,000. In Section II of this paper, we discuss the measurement
of inequality across metropolitan areas.
Just as at the national level, inequality across metropolitan areas reflects the distribution
of human capital, the returns to human capital and governmental redistribution. A
primary difference between local and national level inequality is that local inequality is
driven to a large extent by decisions of people to live in different places. According to
2006 American Community Survey, seven percent of the U.S. population lives in a
different county or country than they did only one year ago, and twenty-one percent of
3
the population lives in a different county or country than they did five years ago.
Manhattan’s inequality reflects the decisions of both rich and poor to come to the island.
Since more than fourteen percent of Kendall’s population lived outside the county one
year ago, the area’s reflects the fact it attracts homogenous people.
Paradoxically, local inequality is actually the inverse of area-level income segregation.
Holding national inequality constant, local inequality falls as people are stratified across
space so that rich live with rich and poor live with poor. A perfectly integrated society,
where rich and poor were evenly distributed across space, would have highly unequal
metropolitan areas that mirror the entire U.S. income distribution.
In Section III, we find that almost one-half of the variance in income inequality across
space can be explained by differences in the skill distribution across metropolitan areas.
Places with abundant college graduates and high school dropouts are areas that are
particularly unequal. Traditional economic models try to explain the location of skilled
and unskilled workers with differences in the returns to skill and differences in amenities
(Dahl, 2002). We agree with this framework, but empirically, we find that history and
immigration seem to be the most important determinants of inequality today.
Sixty percent of the heterogeneity in skills across larger metropolitan areas can be
explained by the share of high school dropouts in the area in 1940 and the share of the
population that is Hispanic. Long-standing historical tendencies are highly correlated
with the location of high school dropouts and the location of Hispanic immigrants today.
Historical skill patterns also play a huge role in the current location of college graduates
(Moretti, 2004), and explain much of the distribution of skills across space.
Metropolitan inequality also reflects differences returns to skill. A modified Gini
coefficient that holds the skill composition of each area constant, but allows the returns to
skill to vary, can explain 50 percent of the variation in the actual Gini coefficient across
metropolitan areas. The correlation between the raw Gini coefficient at the metropolitan
area level and the estimated returns to a college degree in that area is 73 percent.
4
We do not fully understand why some places reward skill so much more strongly than
others. In our data, the most powerful correlate of the returns to having a college degree
is the share of population with college degrees, but that fact provides more confusion
than clarity. The correlation could reflect skilled people moving to areas where there are
large returns to skill. Alternatively, it might reflect human capital spillovers which cause
the returns to skill to rise. Hopefully, future research will help us better understand the
differences in the returns to skill across metropolitan areas.
In Section IV of this paper, we turn to the consequences of local inequality. We find a
significant negative correlation between local economic growth and income inequality
once we control for other initial conditions, such as the initial distribution of skills and
temperature. Places with unequal skills actually grow more quickly, but places with
more income inequality, holding skills constant, have slower income and population
growth. Inequality is related to crime at both the national and city level (Fajnzylber,
Lederman and Lloayza, 2002; Daly, Wilson and Vasdev, 2001). Our data also confirms a
robust relationship between the murder rate and inequality. Luttmer (2005) documents
that people are less happy when they live around richer people. We also find people who
live in more unequal countries report themselves to be less happy.
In Section V, we discuss the policy issues surrounding local inequality. Even if we
accept that local inequality has some unattractive consequences, the consequences of
reducing local inequality, holding national inequality fixed, are quite unclear. Reducing
local inequality, leaving the national skill distribution untouched, implies increased
segregation of the rich and the poor.
Moreover, easy migration across areas severely limits the ability of localities to reduce
income inequality through redistributive policies (Peterson, 1981). The ability to
migrate means that when localities take from the rich and give to the poor, they will
induce the rich to emigrate and attract more poor people, which in turn creates added
burdens on the city’s finances. After all, the evidence in Section III suggests that the
5
location of high skilled people may be quite sensitive to the returns to skill. Decreasing
the returns to skill at the local level with higher taxes or other policies will surely induce
some skilled workers to leave.
Our final point is that America’s current schooling system puts localities at the center of
any attempts to reduce national inequality through a more equal human capital
distribution. While localities are inherently weak in their abilities to reduce inequality,
decentralized schooling means that any attempt to equalize educational opportunities
must rely heavily on localities. Not only will attempts to reduce inequality through
more equal education take many years, but they will also require a tricky partnership
between national and local governments.
II. Measuring Inequality Across American Metropolitan Areas
In the analysis that follows, we will use metropolitan areas as our geographic unit and the
Gini coefficient as our measure of income inequality. Metropolitan areas are defined as
multi-county agglomerations that surround a city with a “core urban area” of over 50,000
people. Metropolitan areas have the advantage of at least approximating local labor
markets, and they are large enough to provide a certain measure of statistical precision.
The disadvantage of using these areas is that they do not correspond to natural political
units, which makes them awkward units for analyzing or discussing public policy.
Much of the data for this paper comes from the five-percent Integrated Public-Use Micro-
Samples (IPUMS) for the 1980 and 2000 Censuses (Ruggles et al, 2008). In most cases,
we will restrict ourselves to total household income, which means that we are not treating
single people differently from married people. We will focus on pre-tax income
inequality.1 A particular problem with using Census data to measure income inequality is
that incomes in the 2000 Census are top-coded at 999,998 dollars and at 75,000 in 1980.
1 Our income measures, like many in this literature, exclude non-income returns from capital ownership, like the flow of services associated with owning a home.
6
In the case of top-coding, we use the income top-code, but recognize that this is
understating the true degree of income inequality.
We use the Gini coefficient as our measure of inequality, mainly because it is the
ubiquitous standard in the inequality literature. Many policy discussions of inequality
focus on poverty, and reducing poverty levels is a natural topic of policy attention. Yet
this essay is focused on extremes at both the upper and lower levels of the income
distribution, and the Gini coefficient captures that heterogeneity. The Gini coefficient,
defined as ∫ −−y
dyyFy
2))(1(ˆ11 , where y is the mean income in the sample and F(y) is
the share of the population with income levels less than y. This measure has the
interpretation as the area between the 45 degree curve (which indicates perfect equality)
and the Lorenz curve.2
The Gini coefficient has the advantage of being invariant with respect to scale, so that
larger areas or richer areas do not necessarily have larger or smaller Gini coefficients.
Moreover, a ten percent increase in everyone’s income will not impact the Gini
coefficient. The Gini coefficient also always rises when income is transferred from a
poorer person to a richer person. One standard criticism of the Gini coefficient is that the
average Gini coefficient of a number of areas will not equal the Gini coefficient
calculated for those areas all together.
There are several plausible alternatives such as the variance of income within an area (
∫ −y
ydFyy )()ˆ( 2 ) or the coefficient of variation ( ∫ −y
ydFyyy
)()ˆ(ˆ1 2 ). The
coefficient of variation, unlike the variance, will not increase or decrease if all incomes
are scaled up or down by the same percentage amount. A final type of measure is the
difference in income between individuals in different places of the income distribution, 2 If we let p denote F(y), i.e. the share of the population earning less than y, then the Lorenz curve plots the share of national income going to individuals earning less than )(1 pF − as a function of p, i.e.
∫ −≤ )(1)(
ˆ1
pFyydyyf
y.
7
for example the difference in income between people at the 90th and 10th, or the 75th and
25th, percentiles of the income distribution. We calculate these differences using the
logarithm of income.
Using the 2000 Census five percent micro-sample, we calculate these five different
income inequality measures at the metropolitan area level: (1) the Gini coefficient using
household income, (2) the variance of household income, (3) the coefficient of variation
of household income, (4) the income difference between the 90th and 10th percentiles of
the household income distribution, calculated as the difference in the logs of these
numbers and (5) the income difference between the log 75th and 25th percentiles of the
household income distribution. Table 1 shows the correlation between our five measures
as well as the correlation between these measures and the logarithm of both median
family income in the area and population size.
The most reassuring fact in the table is that these income measures are fairly highly
correlated. For example, the correlation coefficient between the Gini coefficient and the
coefficient of variation is 92 percent. The correlation coefficient between the Gini
coefficient and the 90-10 percentile income difference is 91 percent. The variance is less
correlated with these other measures because it is highly correlated with the mean level of
income in the area. In general, these different measures give us a similar picture of which
metropolitan areas within the U.S. are most unequal.
We are interested not only in the stability of income inequality between different
measures, but also in the stability of income inequality over time. Figure 2 graphs the
Gini coefficient for 242 metropolitan areas estimated from the 1980 Census against the
Gini coefficient from the 2006 American Community Survey, the most recent data
available. For comparison, we also plot the 45 degree line.
Overall, the correlation between the Gini coefficient in 1980 and the Gini coefficient in
2006 is .58, which suggests neither extreme permanence nor enormous change in the
rankings. Places that had an unusually high level of income inequality in 1980 revert
8
slightly to mean and have relatively less inequality today. As the Gini coefficient in
1980 increases by .1, the growth in the Gini coefficient over the next 26 years falls by
.03. Some of the impermanence (and mean reversion) surely reflects measurement error
in the Gini coefficient.
The most striking fact in Figure 2 is that the points in the graph are above the line in all
but one area (Ocala, Florida), which means that for almost all the MSAs the estimated
Gini coefficient is much higher today than it was 26 years ago. Much of this surely
reflects the real increase in inequality in this country that has been extensively
documented (e.g. in Katz and Murphy, 1992, and subsequent literature). However, some
of the seeming increase in Gini coefficients may reflect changing top codes, but when we
look at other measures that are less subject to top-coding issues (i.e. the 90-10
differences) we continue to see large increases in inequality in almost all areas.
In 1980, the Gini coefficients ranged from .33 to .45. Wisconsin had the most equal
metropolitan areas 25 years ago with the Appleton-Oshkosh-Neenah MSA’s Gini
coefficient of .33, and Gainesville, Florida, was the most unequal metropolitan area with
a Gini coefficient of .45. Two very poor Texas areas (Brownsville and McAllen) had the
next highest levels of inequality. Wisconsin still has the country’s most equal
metropolitan area in 2006 (Sheboygan with coefficient of .38), but even it is substantially
less equal than the most egalitarian places were 25 years ago. New Haven-Bridgeport-
Stamford, with its combination of inner-city poverty and hedge-fund entrepreneurs, is
now the most unequal metropolitan area in the country with a Gini coefficient of .54.
While the county of Manhattan is more unequal, there are no other metropolitan areas
that are even close to that Connecticut area in income inequality. The next three most
unequal areas are Gainesville, Florida; Athens, Georgia; and Tuscaloosa, Alabama.
Inequality shows up both in America’s richest metropolitan areas, like New Haven, and
in some of its poorer areas.
Generally, there is a negative association between area inequality and average incomes.
For example, the Gini coefficient, coefficient of variation, the 90-10 percentile income
9
difference and the 75-25 percentile income difference are all are negatively associated
with area income. Figure 3 shows the –14 percent correlation between the Gini
coefficient and the logarithm of median family income.
However, this connection has been declining over time. Figure 4 shows the -59 percent
correlation between the Gini coefficient and the logarithm of median family income in
1980. This correlation is far stronger than the -14 percent correlation for 2006 shown in
Figure 3. 25 years ago almost all rich places were relatively equal, given the relative
inequality of the United States. Today, some of America’s richest places are also among
the most unequal. Some of this change may reflect changing top-coding, but it surely
also reflects the enormous gains in wealth at the top end of the income distribution over
the past 25 years in wealthy cities like San Francisco.
While the link between average income and inequality is becoming weaker, the link
between area population and inequality is becoming stronger. In 1980, the raw
correlation between population and the Gini coefficient was essentially zero (-.02). In
2000, the Gini coefficient’s 15 percent correlation with area population is shown in
Figure 5. Regressions (1) and (2) in Table 2 show bivariate regressions where the Gini
coefficient is regressed on contemporaneous income and population measures in 1980
and today. Between 1980 and 2000, the connection between average income and the
Gini coefficient fell by more than 40 percent, and the connection between area population
and the Gini coefficient increased by roughly the same percentage.
Does nominal income inequality imply inequality of real incomes or of consumption?
Prices differ across metropolitan areas, but if prices were the same for every type of
person in every area, then prices should not impact inequality, at least as measured by the
coefficient of variation or the Gini coefficient. However, as suggested by Black,
Kolesnikova and Taylor (2007), prices may be quite different for people at different
places in the income distribution. New York may be much more expensive for a
relatively rich person than it is for a relatively poor person. Indeed, the very fact that
10
poor people continue to live in New York suggests that the area may not be as expensive
for them as average prices would indicate.
This issue could be addressed by developing different price indices for people at different
levels of the income distribution in different metropolitan areas, but that is far beyond the
scope of this paper. Instead, we have undertaken the far simpler task of asking about the
inequality of consumption of one important good: housing. If places with more rich
people are expensive places for the rich to live, then we should expect to see less
inequality of housing consumption than inequality of income.
To calculate a housing consumption Gini coefficient, we must first calculate a measure of
housing consumption for everyone in the U.S. We start with a national housing price
regression, where the logarithm of housing price is regressed on the characteristics of
every household. Because of the limited number of housing characteristic data available
from the Census microsample, we instead use data for 46 of the largest metropolitan
areas from the American Housing Survey Metropolitan Samples for 1998, 2002, 2003
and 2004. Housing characteristics include interior square footage, exterior square
footage, the number of bathrooms, the number of bedrooms and several other features.
We then use this regression to form a predicted housing price measure for every
household. Essentially, we are using a hedonic regression to create a housing price index
that enables us to aggregate across different housing characteristics.
We use this housing consumption measure to calculate a Gini coefficient of housing
consumption for every metropolitan area. Figure 6 shows the 38 percent correlation
between this housing consumption Gini coefficient and our income Gini coefficient.
More unequal incomes also have more unequal housing consumption, but in general
housing consumption inequality is much less than income inequality and housing
consumption inequality is particularly below income inequality in places with large
amounts of income inequality. The mean housing consumption Gini coefficient is 0.28,
much lower than 0.45, which is the mean of the household income Gini for this
subsample of metropolitan areas in 2000.
11
While some of this difference can be attributed to measurement error, since our hedonic
regression omits many key housing attributes, this result still suggests that places with
highly unequal income levels have less housing consumption inequality than one might
expect. This fact does not prove that prices are largely offsetting incomes, but
heterogeneity in local prices that impact rich and poor people differently may be
important, and measuring these prices and their impact is yet another interesting topic for
future research.
III. The Causes of Urban Inequality
The typical economic approach to earnings is to assume that they reflect the interaction of
human capital and the returns to human capital (e.g. Katz and Murphy, 1992). Indeed,
human capital is often defined as including everything that goes into earnings, in which
case the relationship is essentially tautological. If human capital is reduced to being a
scalar, h, then the wage associated with each value of h is )(hwi where i represents each
place. If the density of population in each area with human capital level h is )(hgi , then
the average earnings in a locality is equal to ∫h ii dhhghw )()( . The density of income
will be ))(( 1 ywg ii− where ))(( 1 ywG ii
− denotes the cumulative distribution of income.
If hhw iii βα +=)( , then the variance of wages within a place is equal to )(2 hVariiβ ,
where )(hVari is the variance of h within place i. The coefficient of variation is
)(ˆ hVarh
iiii
i
βαβ+
, where ih is the mean of h within place i. The Gini coefficient is
∫ −−−y ii
i
dyywGy
21 )))((1(ˆ11 , and if h is distributed uniformly on the interval
[ ]iiii hh σσ 5.ˆ,5.ˆ +− then the Gini coefficient is ( )iii
ii
h3)(
13
βασβ
+− , which is a function of
both the distribution of skills and the returns to skill.
12
The inequality of after-tax, after-redistribution earnings will then also be affected by the
progressivity of the tax rate and the state of the social safety net. We will turn to the
heterogeneity in welfare payments later, but our primary focus is on the dispersion of
before-tax earnings. We will begin by discussing the role that heterogeneous human
capital plays in explaining the differences in inequality across space and the causes of
that heterogeneous human capital. We will then turn to differential returns to human
capital and governmental after-tax redistribution.
Human Capital Heterogeneity and Income Inequality Across Areas
To assess the role that human capital plays in explaining income inequality across areas,
we will take two complementary approaches. First, we will simply regress the Gini
coefficient on measures of human capital. Second, we will create Gini coefficients for
each metropolitan area based on the observable measures of human capital and national
wage regressions. Both measures are compromised by the fact that our measures of
human capital are coarse. They capture only the years of formal schooling and years of
experience. True human capital would include many more subtle factors, such as the
quality of schooling and of experience.
In regression (3) of Table 2, we examine the relationship between the Gini coefficient in
2000 and two primary measures of human capital in the same year: the share of adults
with college degrees and the share of adults who are high school graduates. We continue
to control for area population and area income. Both of these variables are extremely
significant and they increase the amount of variance explained (i.e. r-squared) from 15
percent to 49 percent. As such, more than one-third of the heterogeneity in income
inequality across metropolitan areas can be explained by these two basic measures of
human capital.
The coefficients are also large in magnitude. As the share of college graduates increases
by 10 percent, the Gini coefficient rises by .031, a little more than one standard deviation.
13
As the share of high school graduates increases by 10 percent, the Gini coefficient drops
by .018, about two-thirds of a standard deviation. While it may be unsurprising that
even such crude proxies for heterogeneity in the human capital distribution can do so well
at explaining the income distribution, this fact illustrates that much of inequality within
an area reflects the heterogeneity of skills within that area.
One concern about results such as these is that perhaps the inequality of human capital
within an area is itself an endogenous response to changes in the returns to skill. If places
that have high returns to having a college degree attract people with college degrees, then
controlling for the skill distribution in this way may also end up controlling for the
returns to skill. After all, economic theory predicts that college graduates should go to
places where the returns to being a college graduate are higher. Of course, this cannot
explain why inequality is higher where there are more high school dropouts, but it is still
worth taking the endogeneity of skills seriously.
One approach to this endogeneity is to look at long-standing historical skill patterns. We
only have data on the share of the population with high school and college degrees going
back to 1940. In regression (4), we show the results using those variables instead of
contemporaneous college and high school graduation levels. We continue to control for
contemporaneous income and population, but the results are unchanged if we remove
those controls. Human capital levels in 1940 are still strongly correlated with inequality
today. The overall r-squared declines to 32 percent, but these historical variables
incrementally explain more than fifteen percent of the variation in the Gini coefficient.
The coefficient on the high school graduation rate in 1940 is quite close to the coefficient
on 2000 high school graduation. The coefficient on the college graduation rate is more
than three times higher using the older data. However, the variation in the college
graduation rate is much smaller in 1940, so a one standard deviation increase in the
college graduation rate has about the same effect when using 1940 or 2000 data. One
way to understand why 1940 college graduation rates have such a strong impact on
14
modern inequality is that they strongly predict the growth in college graduation rates after
that point.
In the fifth regression, we go back even further and look at 19th century measures of
human capital. We do not have measures of human capital in the adult population, but
we do have enrollment rates for high schools and colleges during this time period, which
are found in historical Census data.3 Unfortunately, we lose over 70 metropolitan areas
by using this historical data. The coefficients on these enrollment rates are not, therefore,
particularly comparable with the coefficients on population-based skill measures in
regressions (3) and (4). In regression (5), we find that the share of the population
enrolled in college in 1850 is a quite solid predictor of income inequality today. High
school enrollment rates also continue to negatively predict inequality.
In this case, the incremental r-squared created by those two variables when compared to
regression (2) is quite modest (five percent). Still, we are impressed by the ability of
150-year-old educational variables, measuring something quite different from modern
skill levels, to explain anything. The coefficient on college enrollment in 1850 is quite
large, but again this needs to be considered together with the extremely small level of
variation in this variable across space. These results continue to suggest to us that
historical patterns of human capital play some role in explaining income inequality today.
In the sixth regression, we add two more historical measures of human capital: (1) the
share of the population that is illiterate in 1850 and (2) the share of the population that
was enslaved in that year. We interpret both measures as proxies for human capital
deprivation. Learning to read is an obvious measure of human capital. Slaveowners
often opposed education for their slaves. Indeed, during the century after emancipation,
the former slave areas continued to provide particularly poor education for African-
Americans. Both illiteracy and slavery in 1850 help predict inequality today. Adding
3 Haines, M.R., (2004) ICPSR study number #2896, Historical Demographic, Economic, and Social Data: The United States, 1790-2000.
15
these variables increases the r-squared of the regression to 26 percent, an additional six
percent.
Another way of looking at the impact of human capital is to ask whether human capital in
1980 is associated with growth in inequality after that year. In Table 3 we regress the
Gini coefficient in the year 2006 on the Gini coefficient in the year 1980 and other
controls. We use the 2006 measure, rather than the 2000 measure, because it increases
the time period of change by 30 percent. Regression (1) can be interpreted as a growth
regression since we are asking about the determinants of inequality today, holding past
inequality constant. The coefficient of .79 on the Gini coefficient in 1980 implies that
there is mean reversion between 1980 and today, although this could be the result of
measurement error. The result on income and population tell us that income inequality
has been rising both in bigger areas and in richer areas.
In regression (2), we include our controls for human capital in 1980. Again, there is a
strong positive association between college graduation rates in 1980 and inequality today.
Places with more highly skilled people in 1980 have become more unequal over time,
which presumably reflects both the rise in the returns to skill and the tendency of skilled
people to move to more skilled areas, which we will discus later. Places with more
college dropouts have also become more unequal over time. Not only do
contemporaneous skill levels predict inequality, but inequality of skills in 1980 predicts
an increase in income inequality since then.
We now turn to our second means of assessing the importance of skill distributions in
explaining the inequality of income. In this approach, we calculate only the income
inequality from males between the ages of 25 and 55. To keep sample sizes up, we look
only at the 102 metropolitan areas with more than 500,000 people. We use only workers
with positive earnings, and we use only labor market income. We calculate three Gini
coefficients. First, we calculate the standard Gini coefficient using the earnings from
these workers. Figure 7 shows the 74 percent correlation between this Gini coefficient
and our household income Gini coefficient among these 102 metropolitan areas.
16
We then compare this Gini coefficient for male workers with Gini coefficients based
entirely on the human capital of these workers, by which we mean age and years of
schooling. To calculate these “human capital only” Gini coefficients, we use a
nationwide earning regression to predict earnings for everyone in the sample. By using
these predicted earnings rather than true earnings we can isolate the impact of the level of
human capital in an area while abstracting away from the differential returns to
schooling. We calculate the Gini coefficient based on these predicted earnings. Figure 8
shows a 57 percent correlation between the two Gini coefficients.
Our Gini coefficient based only on human capital explains about 33 percent of the
variation in overall income inequality among working-age males. This is a considerable
amount of explanatory power, but it still leaves plenty to be explained. Another way of
looking at the data is to note that the average Gini coefficient calculated with human
capital only is one half the size of the Gini coefficient calculated using true income. As
such, human capital gets you some, but far from all, of the way towards explaining the
amount of income inequality.
One reason why these human capital based measures are failing to more fully explain
actual income inequality might be that our human capital measures are so coarse.
Occupations may provide us with a richer means of measuring individual-level human
capital. As such, we created a third Gini coefficient using the same sample but including
occupation dummies in our wage regressions. These dummies are then used, along with
years of schooling and age, to predict wages, and these predicted wages are used to
calculate a local Gini coefficient. The mean of this occupation-based Gini coefficient is
about half-way between the Gini coefficient with just education and age and the Gini
coefficient of true income. If we accept that occupation is a measure of skills, then this
measure goes much more of the way towards explaining income inequality across areas.
Figure 9 shows the 86 percent correlation between this occupation-based Gini coefficient
and the true income Gini coefficient. The occupation-based index has the ability to
17
explain 73 percent of the variation in the true income Gini coefficient. There are two
interpretations of this finding. First, occupation may be proxying for income more than
human capital and therefore should not be used as a control in the regression. Second,
occupation may indeed be a better measure of human capital. There is surely some truth
in both views, but we can draw the conclusion from this section that heterogeneity in
human capital across space can explain a considerable amount of the heterogeneity in
income inequality across space.
The Causes of Human Capital Inequality
Why are the levels of human capital so different in different places? One theory is that
current human capital levels reflect long-standing education policies formed over the last
200 years; Goldin and Katz (2008) provide a rich description of this history. Urban and
economists who emphasize people’s location decisions will tend to focus on differences
in economic productivity and differences in amenities that then motivate migration (e.g.
Dahl, 2002). According to this view, areas that specialize in industries which are
particularly good for low or high skill workers should then disproportionately attract
those workers. Of course, this theory just pushes the puzzle one step backward. A more
complete explanation for heterogeneity in skills would also explain why different
industries are located in different places. In some cases, like cities with ports or coal
mines, there are exogenous factors that explain industrial location, but observable
variables tend to only explain a modest amount of industrial concentration (Ellison and
Glaeser, 1999).
Alternatively, amenities may draw high-skill workers to a particular location. For
example, if there is some amenity that is particularly desirable and in particularly short
supply, then we would expect rich people to locate in places with that amenity. This can
certainly explain why there are so many rich people in Paris (Brueckner, Thisse and
Zenou, 1999) or on the Riviera. Alternatively, there can be other amenities, such as
access to public transportation, which might draw poor people disproportionately to a
given area.
18
While the economic framework that emphasizes rational location choice certainly has
some ability to explain the distribution of skilled people across space, much of current
skill patterns appear to be determined by long-standing historical skill patterns, as
discussed above. For example, Figure 10 shows the 73 percent correlation between the
share of adults with college degrees in the year 2000 and the share of adults with college
degrees in the year 1940. The college share of the population in 1940 is able to explain
more than 50 percent of the variation in the college share today, which suggests the
enormous power of historical forces in shaping the skill composition of cities today.
We also run a regression that shows that skill growth is strongly predicted by the initial
skill level. As the 1940 share of the adult population with college degrees increased by 5
percent, the growth rate in the share of the population with college degrees between 1940
and 2000 increased by 10 percent. Far from there being mean reversion in this variable,
there has been a tendency of skill growth to be concentrated in places that began with
more skills (Berry and Glaeser, 2005).
The relationship between today and the past is much weaker at the bottom end of the skill
distribution. Figure 11 shows the 45 percent correlation between the share of the adult
population who did not have a high school degree in 1940 and the share of the adult
population without a high school degree in 2000. In this case, the 1940 variable can only
explain one-fifth of the variation in the high school dropout rate today. The graph shows
a number of places, such as Miami and McAllen, Texas, which have particularly large
high school dropout rates today relative to their historical levels. Older variables, such as
the percent of the population enslaved in 1850, can explain about one-tenth of the
variation in the high school dropout rate today.
These outliers suggest that immigration, particularly from Latin America, is also
associated with a heavy concentration of less skilled workers. Figure 12 shows the 57
percent correlation between the share of the population that is Hispanic and the share of
the population without a high school degree in 2000. Together, the Hispanic share today
19
and the dropout rate in 1940 can explain 61 percent of the variation in the dropout rate
today. The Hispanic share is particularly high in Florida, Texas and California, which are
three states that were once part of the Spanish Empire and which are geographically close
to Mexico and other countries in Latin America. In fact, the correlation between
Hispanic share and MSAs latitude is -36 percent, which reinforces this point. Once
again, history seems to play a large role in explaining the current skill distribution.
By contrast, the evidence supporting the importance of the traditional economic
explanations of the location of talent is much weaker. For example, there is little
evidence that highly skilled people have moved to areas with particularly pleasant
temperatures. There is a no robust association between January temperature and the
share of the population with college degrees. High July temperatures are associated with
fewer college graduates, but even this effect is quite modest. July temperatures can
explain only 6.5 percent of the variation in the share of the population with college
degrees.
LeRoy and Sonstelie (1983) and Glaeser, Kahn and Rappaport (2008) provide theory and
evidence supporting the view that less skilled people live in the centers of metropolitan
areas because of access to public transportation. Cars are expensive, and poorer people
prefer the time-intensive, lower-cost alternative of buses and subways. Can access to
public transportation explain the location of less-skilled people across areas? No. There
is virtually no correlation across metropolitan areas between the share of the population
without high school degrees and the share of the population that takes public
transportation. This absence of correlation is particularly surprising since the poor
generally take public transit more, which should yield a positive relationship between less
skilled people and public transit, even if the poor didn’t move across metropolitan areas
in response to public transit access.
Alternatively, people of different skill levels may be drawn to particular areas because of
skill-specific economic opportunities. Silicon Valley has a booming computer industry,
and it attracts extremely highly skilled engineers. New York City attracts smart people to
20
work in finance. Certainly, there is a strong correlation between the skill level of an area
and the skill orientation of the industries in the area. Using the 2000 Census, Glaeser
and Gottlieb (2008) ranked industries by the share of the workers in that industry in the
nation with a college degree. They then calculated the share of a metropolitan area’s
employment that was in the top 25 percent industries ranked by human capital and in the
bottom 25 percent of industries ranked by human capital.
Figure 13 shows the 79 percent correlation between this measure of high skilled
industries and the share of the population with college degrees. Figure 14 shows 49
percent correlation between the share of the adult population without a college degree
and the measure of low-skill industries. Certainly, there is a robust relationship between
the skill orientation of the industries in an area and the skill distribution of the area. But
which way does the causality run? Are skilled industries moving into an area because
there are an abundance of skilled workers, or are skilled workers moving to areas because
of skill-oriented industries?
While surely both phenomena occur, we think that the evidence supports the view that
industries are responding to the area’s skill distribution more than the view that the skill
distribution is responding to the area’s industries mix. For example, the share of the
population with college degrees in 1940 can explain 35 percent of the variation in the
skill mix of industries today. By contrast, the skill composition of the industries in the
metropolitan area in 1980 can only explain seven percent of the variation in growth of the
population with college degrees since that date. The complex two-sided nature of this
relationship makes it difficult to accurately assess the direction of causality, but there are
reasons to think that much of the industrial mix in the area is actually responding to the
skill distribution.
One variable that seems more plausibly predetermined is the concentration in
manufacturing during the first half of the 20th century. The location of factories does not
seem likely to be particularly driven by the presence of highly skilled workers 100 years
ago. Yet an industrial orientation, as late as 1950, is negatively correlated with the share
21
of the population with college degrees today, perhaps because those manufacturing cities
tended not to reinvent themselves as centers of idea-oriented industries or perhaps
because manufacturing employers were less disposed towards high schools earlier in the
century (Goldin and Katz, 2008). As the share of the workforce in manufacturing in
1950 increases by 10 percent, the share of the population with college degrees drops by
about 1 percent. This may explain why, as shown in Table 2, Regression (7),
manufacturing in 1950 is negatively associated with inequality today.
These industrial measures are essentially proxies for differential returns to human capital
across metropolitan areas. Using wage regressions, we are able to estimate such
differential returns directly by running regressions of the form:
(1) ControlsOtherHSGradBAGradWageLog HSMSA
BAMSAMSA +++= **)( ββα
where MSAα is an area specific intercept, BAMSAβ is an area specific return to having a
college degree and HSMSAβ is an area specific return to having a high school diploma. This
regression estimates a differential return to different levels of schooling for each
metropolitan area. We estimate this regression only for prime age males, and include
controls for experience. We focus only on those areas with more than 500,000 people so
that the returns to schooling are estimated with reasonable precision.
Figure 15 shows the 24 percent correlation between our estimate of BAMSAβ and the share of
the adult population with college degrees in the 102 metropolitan areas in our sample
with more than 500,000 people. One interpretation of this fact is that skilled people are
moving to places where the skill levels are higher. A second interpretation is that
agglomerations of skilled people raise the returns to skill. Any interpretation of this
relationship is compromised further by the fact that an abundance of skilled people would
normally reduce the returns to skill in a typical model of labor demand. Still, this does
suggest that highly skilled people are living in places where the returns to skill are higher.
22
Differential Returns to Human Capital
The fact that the returns to capital differ across space can also potentially explain the
inequality that we see across metropolitan areas. As the formula discussed above
illustrates, pre-tax income inequality will reflect both differences in the distribution of
skills and differences in the returns to those skills. Certainly, the framework predicts a
strong link between places with higher returns to college and income inequality.
Figure 16 shows the 73 percent correlation between our estimated return to a college
degree and the Gini coefficient across the 102 areas with more than 500,000 people. The
measured return to a college degree is much better at explaining area inequality than the
number of people with college degrees. Of course, these returns are directly based on the
same income data that is being used to generate the Gini coefficient. Still, this finding
seems to confirm the view that heterogeneity in returns to skill can help us to explain
differences in income inequality across space.
To look at this further, we again calculate Gini coefficients for each metropolitan area
using wage regressions. However, in this case, we allow the coefficients on skills to
differ across metropolitan areas as shown in equation (1) above. We again run these
regressions only for prime aged males. We then use these regressions to predict the
amount of inequality in an area if the skill distribution of the area were the same as the
skill distribution in the country as a whole. When we calculated Gini coefficients using
wage regressions above, we were calculating local Gini coefficients based only on
differences in the skill composition, holding the returns to skill constant across space.
Now we calculate Gini coefficients holding the skill composition constant, but allowing
the returns to skill to differ across space.
Figure 17 shows the 71 percent correlation between these predicted wage Gini
coefficients and actual Gini coefficients in our sample of prime age males across
metropolitan areas with more than 500,000 people. The relationship is tighter than it was
when we looked at Gini coefficients that assumed a constant return to skill. Moreover,
23
this Gini coefficient holding the skill composition constant explains 50 percent of the
variation in the actual Gini coefficient, whereas our constant return to skills Gini
explained only 33 percent of the difference. We interpret these results as suggesting that
differential measured returns to human capital can explain area-level income inequality
somewhat better than differences in measured human capital.
One potential concern with interpreting these results is that measured returns to human
capital may not be measuring higher returns to human capital, but instead measuring high
levels of true human capital associated with each coarse category of observed human
capital. For example, if people with college degrees in some areas went to higher quality
schools or have had better work experience, then this would cause the measured return to
a college education to increase, even if the true returns to human capital were constant
across space. We have no way of dealing with this hypothesis, and we will continue
referring to the measured returns to human capital as the returns to human capital,
understanding that it can also reflect other things.
While differences in the returns to skill do seem to explain a significant amount of the
differences in inequality, we do not know what explains differences in returns to skill
across space. For example, the positive correlation between the share of the population
with college degrees and returns to skill might suggest that being around other skilled
people increases the returns to being skilled. Alternatively, Beaudry, Doms and Lewis
(2006) suggest that places with abundant skilled workers invested in computerization,
which then had the effect of raising the returns to skill. While we are certainly
sympathetic to these interpretations, it is hard to distinguish between this view and the
view that more skilled people are moving to areas where the returns to skill are higher.
Moreover, the share of the population with college degrees in 1940 does little to explain
the returns to college today. If we thought that higher returns to skill reflected the power
of agglomerations of skilled people, then an abundance of skills in 1940, which predicts
skills today, should also predict higher returns to college today. This is not the case, as
we find only a 5 percent correlation between the share of the population with a college
24
degree in 1940 and our measure of the return to a college degree. Instead, a higher return
to college today is strongly associated with recent growth in the share of the population
with college degrees. The correlation between our return to college measure and the
change in the proportion of the adult population with college degrees is 26 percent.
These facts support the idea that skilled people are moving to areas where the returns to
skill are higher.
Other variables also do a relatively poor job of explaining the returns to skill. For
example, our measure of skill intensive industries doesn’t explain the returns to skills.
Looking at Figure 15 shows that some of the places with the highest returns to skill are
usual suspects. The financial agglomeration in Southwestern Connecticut and the
technology agglomeration in San Francisco Bay have very high returns to human capital.
But there are also areas, like Houston and Birmingham, that are more of a surprise.4
Figure 18 illustrates the 34 percent correlation between our estimated returns to college
and the share of workers in finance among the 102 cities in our sample with more than
500,000 people.5 Figure 19 shows the 27 percent correlation between the returns to
college and the share of workers in the computer industry among the same sample of
cities.6 This seems to support the results of Beaudry, Dom and Lewis (2006) who show a
link between computerization and inequality. A related and interesting hypothesis is that
wage inequality is linked to the former concentration of low skilled workers in routine
tasks that have now been made obsolete (Autor and Dorn, 2007).
Still, we are much more confident that differences in the returns to skill can explain a
significant amount of income inequality across metropolitan areas than we are in
explaining why areas have such different returns to human capital. A number of recent
papers, like Autor and Dorn (2007); Black, Kolesnikova and Taylor (2007); and Beaudry, 4 Like Black, Kolesnikova and Taylor (2007), we find a modest positive relationship between cost of living and returns to college. 5 “Finance” is defined using IPUMS 2000 Occupation Codes for the 5% sample at http://usa.ipums.org/usa/volii/00occup.shtml. Finance codes are 12, and 80-95. 6 “Computers” is also defined using IPUMS 2000 Occupation Codes for the 5% sample at http://usa.ipums.org/usa/volii/00occup.shtml. Computer codes are 11, 100-111, and 140.
25
Dom and Lewis (2006) have brought some understanding to this question, but it remains
a pressing topic for future research.
IV. The Consequences of Urban Inequality
At the national level, income inequality has been linked to low levels of economic
growth, perhaps because inequality leads to political strife (Persson and Tabellini, 1994;
Alesina and Rodrik, 1994), but local inequality is not the same thing as national
inequality. No one should be surprised if the political and economic effects of inequality
are different at the local level. After all, local political outcomes are far more constrained
by state constitutions and easy out-migration.
More generally, local inequality, as opposed to local poverty, is not necessarily a bad
thing. If people of different income levels mix throughout the country, then local
inequality will be higher than if people segregate into homogenous, stratified
communities. A large number of studies suggest economic mixing, i.e. local inequality,
benefits the less fortunate by giving them more successful role models (Wilson, 1987) or
employers (Mazzolari and Ragusa, 2007). Others suggest that the wealthy develop
empathy for the poor through spatial proximity (Glaeser, 1999). Egalitarians can
simultaneously hope for policies that would reduce inequality at the national level, such
as increasing the schooling levels for least fortunate, while opposing policies that would
reduce local income inequality by moving rich people away from poor people.
Persson and Tabellini (1994) found a strong negative relationship between national
income inequality and economic growth. Some facts about urban growth are quite
similar to facts about country growth. For example, schooling predicts growth at both the
country and the city level. Does the connection between inequality and growth carry
over to the metropolitan level?
In Table 4, we look at the relationship between inequality and growth across our sample
of metropolitan areas. We use 1980 as our start date and look at the growth of both
26
income and population after that year. While country-level regressions typically look
only at income growth, city level growth regressions look at population and income (and
sometimes housing values as well), since increases in productivity should show up both
in higher wages and in more people.
The first regression of Table 4 shows that the raw relationship between income inequality
and local area growth is positive. Places with more inequality have been gaining
population. However, as the second regression in Table 4 shows, this result is not robust
to including a number of other city level controls, such as human capital variables and
January temperature. With these controls, inequality has a significant negative effect on
area population growth. The next two regressions in Table 4 show the relationship
between inequality and area income growth. After including area level controls,
inequality has a significant negative impact on income growth. The coefficients here
should be interpreted while keeping in mind the relatively low level of variation in the
1980 Gini variable. Going all the way from the bottom of the inequality distribution at
0.33, to the top of the distribution at .45, would cause population growth to fall by about
1 standard deviation.
These results do suggest that income inequality is only negatively correlated with area
growth once we control for skills. Increases in the skill distribution that make a place
more unequal by increasing the share of highly educated citizens are associated with
increased, not decreased, growth. However, growth of both income and population was
lower in places where the income distribution is particularly unequal, holding skills
constant.
A second adverse consequence of inequality at the country level is the connection
between inequality and crime (Fajnzylber, Lederman and Lloayza, 2002). These results
have also been found at the metropolitan area level (Daly, Wilson and Vasdev, 2001).
We duplicate them here.
27
In regression (5) of Table 4, we show the strong positive relationship between income
inequality and murder rates across metropolitan areas. We focus on murder rates because
they are the most serious crime outcome and the outcome that is least likely to be
impacted by reporting differences across areas. The 35 percent correlation, shown in
Figure 20 is quite strong. Regression (6) of Table 4 shows that the inequality-crime
relationship is robust to a number of other controls.
Why do murder rates increase with inequality? One view is that inequality is just
proxying for poverty, but both at the country and city level, the impact of inequality on
crime survives controls for the mean income level and the poverty rate. A second
explanation is that inequality leads to less focus on providing community-wide public
goods, like policing. A third explanation is that inequality breeds resentment which then
shows up in higher murder rates.
All of these explanations remain speculation, but there is some evidence that links
unhappiness to envying richer neighbors. Luttmer (2005) looks at the self-reported
happiness of individuals as a function of the wealth of their neighbors. He finds that
people who have richer peers are more likely to say that they are unhappy. The existence
of envy can, under some conditions, suggest that sorting by income is preferable to highly
unequal areas.7
In Figure 21, we show the -47 percent correlation between the Gini coefficient and the
average self-reported happiness in the metropolitan area taken from the General Social
Survey. These happiness data span the last 25 years, and they represent the share of
people who say that they are very happy. Inequality can explain 22 percent of the
variation in this unhappiness measure, and this result is robust to a reasonable number of
other controls such as average area income and population size.
V. Government Policy
7 We have also looked at whether there is a correlation between income inequality and racial segregation, using dissimilarity measures of segregation (see Cutler, Glaeser and Vigdor, 1999 for details of the measure). We find no evidence of any such connection.
28
There are two ways in which government policy interacts with the study of urban
inequality. First, government policy may itself be a cause of that inequality. Second, if
policy makers seek to reduce local inequality, then the study of that inequality may
improve the quality of decision-making. We start with a discussion of the role that
governmental actions might have on the level of inequality and then turn to a discussion
of potential policy implications.
Government Policies and Local Inequality
There are at least three channels through which government policy might impact the
degree of income inequality across space. First, education is largely a government
service, and government policies towards education could either widen or narrow the
distribution of skills within an area. Second, government policies can also impact
migration in ways that might increase or decrease the skill distribution. Third, the
government engages in taxes and redistribution, which would impact the after-tax income
distribution. Some of these policies are explicitly intended to impact inequality, and
other policies are intended to achieve different results but still could end up changing the
level of local inequality.
Investment in school certainly appears to impact the distribution of skills within an area.
For example, Moretti (2004) shows that the presence of a land-grant college in a
metropolitan area prior to 1940 is positively correlated with the skill level of the area
today. When we regress the area Gini coefficient on Moretti’s land-grant college
indicator variable, controlling for area population, income and the share of adults who are
high school dropouts, we find a positive, but statistically insignificant, impact of land
grant colleges on inequality. This effect disappears when we control for share of adults
with college degrees, which implies that this variable (weakly) increases inequality
because it increases the share of more skilled people.
29
Conversely, we find a modest negative relationship between current high school
enrollment rates and inequality when we control for area income and area population.
This result may reflect the tendency of poverty to lead to low enrollment rates or the
tendency of middle income people to move to areas with fewer dropouts or the ability of
high school graduation to reduce inequality. We will not try to distinguish these
hypotheses but just point out that correlations of this form suggest that education policy
can surely impact inequality, both by its direct effect on the skill distribution and by
shifting migration patterns.
There is a long economic literature that suggests that different local level government
policies have the ability to induce selective migration. For example, Borjas (1999)
argues that heterogeneity in welfare policies across space has had a huge impact on the
location patterns of less skilled immigrants, and especially their tendency to locate in
California. Blank (1988) also found that higher welfare levels impact the location
decision of unmarried women with children. This type of effect can explain the poverty
of East St. Louis, which traditionally had higher welfare payments because it lies on the
Illinois side of the Mississippi River within the St. Louis metropolitan area.
Less work has been done on the impact of redistribution on the location decisions of the
rich, but what evidence does exist supports the view that the wealthy are quite mobile and
respond to attempts at redistribution. Certainly, within metropolitan areas, there are good
reasons to think that the rich are sensitive to local tax rates and the bundle of local public
goods. The strong tendency of richer people to live outside of city borders suggests that
they are voting with their feet within certain areas. Haughwout et al. (2004) argue that
these migration tendencies are quite strong and can mean that areas can actually lose
revenue by raising taxes. Feldstein and Wrobal (1998) argue that the migration
elasticities are so strong that states cannot effectively redistribute income at all. This type
of result supports the view that local governments can affect local inequality by moving
people more than they can by classical redistribution.
30
The connection between government policies and local income inequality is in many
ways an understudied topic. We certainly know something about the impact of
education, and we know much about migration responses to local policy differences, but
we do not understand the full contribution that government has played in making some
places more or less equal. For example, local land-use controls that prevent housing for
lower income people can create less inequality within an area, but we do not know how
empirically important this might be.
Local Governments and Local Inequality
Localities do have tools with which they could reduce local income inequality, but it is
not obvious that such tools would enhance welfare. For example, the preceding analysis
suggested that much of the heterogeneity in income inequality across metropolitan areas
was associated with differences in returns to skill. Localities could equalize local
incomes by reducing the returns to skill through more redistributive taxation.
Redistribution would both directly reduce inequality, and is likely to also reduce
inequality by inducing wealthier people to leave the area. However, few localities would
actually find it attractive to increase equality by getting rid of the biggest tax-payers.
While this migration effect might reduce inequality, if it eliminated the richest people in
the city, there are many reasons to think that it would also hurt the area’s economy.
The returns to skill were not closely tied to industrial mix, so trying to attract particular
industries doesn’t seem likely to lead to significant changes in inequality. Moreover, the
historical track record of local industrial policy is decidedly mixed. A long tradition of
urban analysis suggests that localities have a very limited ability to make society more
equal (Peterson, 1981). The ability of wealthy people to flee is just too great.
Greater welfare gains would seem to be associated with policies that enhanced the skills
of the less fortunate. Improvements in school districts and reductions in the size of the
criminal sector could have two possible benefits. First, they might increase the skill
levels at the bottom end of the income distribution. Second, they might attract middle-
31
income people into the area. Local policies that strengthened the bottom of the income
distribution without targeting the top of the income distribution seem most likely to
reduce income inequality without creating other problems.
However, while such policies might well be beneficial, local governments have again
only a limited ability to make the nation-wide skill distribution more equal. Areas with
many poor parents have fewer resources with which to educate their children. These
places have lower tax revenues, holding everything else constant, but they also have less
parental human capital on which to draw. The long-noted power of peers means that
places with lower initial skill distributions inevitably have difficulty creating first rate
public schools.
National Governments and Inequality
Even if one accepts egalitarian ideas that inequality is itself a bad thing, it certainly does
not follow that reducing local inequality is clearly desirable. Would it be sensible for
national government policies to artificially segregate rich people into some cities and
poor people into others? Such segregationist policies would increase equality at the
lower level, but it is hard to see how they would increase local welfare levels. Some of
our regressions have suggested that localities will grow more quickly and have less crime
if they are more equal, but even these results must be treated gingerly. National policies
that created equality by removing high skill, high earnings workers from power areas
would also be likely to reduce the economic performance of those areas. Policies that
reduce the numbers of poor people by eliminating urban attributes that attract those
people, like low-cost housing, would also have negative consequences, most notably the
destruction of a valuable asset that is providing some benefit for the least fortunate
members of society.
At the national level, egalitarianism suggests simple, non-spatial policies, such as classic
income redistribution and policies that support human capital accumulation among the
least fortunate. National policies can also reduce income inequality through
32
redistributive taxation. Those policies involve costs and benefits, and their connection
with cities and localities is relatively modest, since local attempts at redistribution are
likely to create emigration of the wealthy.
Long term reductions in inequality are most likely to be achieved through a more
egalitarian distribution of human capital. Naturally, changes in the distribution of human
capital will take years, if not decades. Those changes will also involve the often
uncomfortable cooperation of national and local governments. Attempts to reduce
inequality by changing the skill distribution must considerably involve localities, given
our current decentralized schooling system. Yet, localities rarely have the resources to
significantly upgrade their schools on their own.
The current structure of local public schooling creates incentives for middle income
people to leave big cities to get better schools for their children. The poor and the very
rich, who send their children to private schools, remain. There could be welfare gains
from an education system that kept the advantages of choice and competition that are
associated with the current system but that also reduced the incentive for middle class
parents to leave big cities.
This fact is the great challenge facing attempts to reduce inequality through schooling.
Our schools are local and localities have a great deal of trouble dealing with inequality.
Poor places have fewer resources to allocate to their schools. Yet if there is going to be a
more equal education distribution, then their schools must be improved. Creating
equality in human capital requires the difficult cooperation of national level education
policy, and schools that often operate at a very local level.
VI. Conclusion
In this essay, we have reviewed the economic causes of metropolitan-area income
inequality. Differences in income inequality across areas can be explained well by both
differences in the skill distribution and differences in the returns to skill. If anything,
33
differences in the returns to skill appear to be more important in explaining the variation
in the Gini coefficient across American metropolitan areas. Differences in the skill
distribution can be well explained by historical tendencies towards having more skilled
people and by immigration patterns. Differences in the returns to skill are far more
difficult to explain, but today the returns to a college degree are highest in areas that
specialize in finance or computing.
There are some negative correlates of area-level inequality. More unequal places have
higher murder rates, and people say that they are less happy. More unequal places grow
more slowly, at least once we control for the skill distribution in an area. The raw
correlation between area-level inequality and population growth is positive.
Area-level income inequality does not create the same policy implications as national
income inequality. At the nation level, an egalitarian, Rawlsian social welfare function
implies the need to reduce income inequality. However, egalitarianism does not provide
the same implications about local inequality. Shuffling people across the country in a
way that creates more homogeneity at the local level would not seem like a natural means
of increasing social welfare given standard social welfare functions. Instead, such
functions would instead push towards a focus on policies like human capital development
that would promote equality nationwide.
We concluded by noting that localities are poorly poised to reduce inequality on their
own. Any attempt at local redistribution is likely to lead to out-migration of the wealthy.
Poor localities don’t have the resources to improve failing schools.
However, if national policies are going to try to reduce inequality by making the
distribution of human capital more equal, then inevitably localities must be involved.
Schools are run at the local level. The combination of national resources and local
operation seems most likely to improve the quality of the poorly performing schools.
Unfortunately, bringing together such different levels of government is inevitably quite
difficult. Moreover, the strong correlation between human capital today and human
34
capital more than fifty years ago suggests that any change will not happen overnight.
35
References
Alesina, A., Glaeser, E. L., 2004. Fighting Poverty in the U.S. and Europe: A World of Difference. Oxford University Press, New York, NY.
Alesina, A., Rodrik, D., 1994. “Distributive Politics and Economic Growth.” Quarterly
Journal of Economics 109(2), 465-490. Autor, D., Dorn, D., 2007. Inequality and Specialization: The Growth of Low-Skill
Service Jobs in the United States, MIT mimeograph. Beaudry, P., Doms, M. E., Lewis, E. G., 2006. Endogenous Skill Bias in Technology
Adoption: City-Level Evidence from the IT Revolution. NBER working paper. Berry, C., Glaeser, E. L., 2005. “The Divergence of Human Capital Levels Across
Cities.” Papers in Regional Science 84(3), 407-444. Black, D., Kolesnikova, N., Taylor, L. J., 2007. Earnings Functions When Wages and
Prices Vary by Location. Federal Reserve Bank of St. Louis working papers. Blank, R. M., 1988. “The Effect of Welfare and Wage Levels on the Location Decisions
of Female-headed Households.” Journal of Urban Economics 24(2), 186-211. Borjas, G. J., 1999. “Immigration and Welfare Magnets.” Journal of Labor Economics
17(4), 607-637. Brueckner, J.K., Thisse, J-F., Zenou, Y., 1999. “Why is central Paris rich and
downtown Detroit poor? An amenity-based theory.” European Economic Review 43, 91-107.
Cutler, D., Glaeser, E. L., Vigdor, J., 1999. “The Rise and Decline of the American Ghetto.” Journal of Political Economy 107, 455-506. Dahl, G., 2002. “Mobility and the Return to Education: Testing a Roy Model with
Multiple Markets.” Econometrica 70(6), 2367-2420. Daly, M., Wilson, M., Vasdev, S., 2001. “Income inequality and homicide rates in
Canada and the United States.” Canadian Journal of Criminology, 219-236. Ellison, G., Glaeser, E. L., 1999. “The Geographic Concentration of Industry: Does
Natural Advantage Explain Agglomeration?” American Economic Review Papers and Proceedings 89(2), 311-316.
Fajnzylber, P., Lederman, D., Loayza, N., 2002. “Inequality and Violent Crime.”
Journal of Law and Economics 45(1), 1-40.
36
Feldstein, M., Wrobel, M.V., 1998. “Can State Taxes Redistribute Income?” Journal of
Public Economics 68(3), 369-396 Glaeser, E. L., 1999. “The Future of Urban Research: Nonmarket Interactions.”
Brookings-Wharton Papers on Urban Affairs, 101-149. Glaeser, E. L., Gottlieb, J. D., 2008. “The Economics of Place-Making Policies.”
Brookings Papers on Economic Activity. Glaeser, E. L., Kahn, M. E., Rappaport, J., 2008. “Why do the Poor Live in Cities? The
role of Public Transportation.” Journal of Urban Economics 63(1), 1-24. Goldin, C. and L. Katz (2008) The Race between Education and Technology.
Cambridge: Harvard University Press. Haines, M. R., 2004. “Historical, Demographic, Economic, and Social Data: The United
States, 1790-2000.” Inter-University Consortium for Political and Social Data, http://www.icpsr.umich.edu/cocoon/ICPSR/STUDY/02896.xml.
Haughwout, A., Inman, R., Craig, S., Luce, T., 2004. “Local Revenue Hills: Evidence
from Four U.S. Cities.” The Review of Economics and Statistics 86(2), 570-585. Katz, L., Murphy, K., 1992. “Changes in Relative Wages, 1963-1987:
Supply and Demand Factors.” Quarterly Journal of Economics 107(1), 35-78. LeRoy, S. F., Sonstelie, J., “1983. Paradise Lost and Regained: Transportation
Innovation, Income and Residential Location.” Journal of Urban Economics 13, 67-89.
Luttmer, E. F. P., 2005. “Neighbors as Negatives: Relative Earnings and Well-Being.”
Quarterly Journal of Economics 120(3), 963-1002 Mazzolari, F., Ragusa, G., 2007. “Spillovers from High-Skill Consumption to Low-Skill
Labor Markets.” IZA Discussion Paper 3048. Moretti, E., 2004. “Estimating the Social Return to Higher Education: Evidence From
Cross-Sectional and Longitudinal Data.” Journal of Econometrics 121(1-2), 175-212. Peterson, P. 1981. City Limits. Chicago: University of Chicago Press. Persson, T., Tabellini, G., 1994. “Is Inequality Harmful for Growth?” American
Economic Review 84(3), 600-621. Ruggles, S., Sobek, M., Alexander, T., Fitch, C. A., Goeken, R., Hall, P. K., King, M., Ronnander, 2008. Integrated Public Use Microdata Series, http://usa.ipums.org/usa
37
Wheeler, C., 2005. “Cities, Skills and Inequality.” Growth and Change 36(3), 329-353. Wilson, W. J., 1987. The Truly Disadvantaged: The Inner City, the Underclass, and Public Policy. Chicago University Press, Chicago, IL.
38
Gini Coefficient
Variance of Household
Income
Coefficient of the Variation of
Household Income
Difference of the Log of Income for
the 90th and the 10th
Percentiles
Difference of the Log of
Income for the 75th and the
25th Percentiles
Log of Median Family Income
Log of Population
Gini Coefficient 1
Variance of Household Income 0.36 1
Coefficient of the Variation of Household Income
0.92 0.20 1
Difference of the Log of Income for the 90th and the 10th Percentiles
0.91 0.29 0.74 1
Difference of the Log of Income for the 75th and the 25th Percentiles
0.86 0.13 0.72 0.93 1
Log of Median Family Income -0.25 0.68 -0.44 -0.17 -0.28 1
Log of Population 0.15 0.57 0.04 0.12 0.00 0.44 1
Table 1Table of Correlations Between Income Inequality Measures for 2000
Source: The Gini coefficient, variance of household income, the coefficient of the variation of household income, the difference of the log of income for the 90th and 10th percentiles and the difference of the log of income for the 75th and 25th percentiles are calculated using the 5% Integrated Public Use Microdata Series (IPUMS) for 2000, at usa.ipums.org. Median family income and population are from the 2000 Census.
39
1980 Gini Coefficient(1) (2) (3) (4) (5) (6) (7)
Ln(1980 Population) 0.0055[0.0012]**
Ln(1980 Median Family Income) -0.1078[0.0084]**
Ln(2000 Population) 0.0081 0.005 0.0075 0.0085 0.0078 0.008[0.0016]** [0.0013]** [0.0014]** [0.0019]** [0.0018]** [0.0015]**
Ln(2000 Med. Family Income) -0.0625 -0.1054 -0.0773 -0.0602 -0.0392 -0.0525[0.0096]** [0.0112]** [0.0091]** [0.0131]** [0.0139]** [0.0096]**
2000 Pct. Of 25+ Pop. With BA 0.3122[0.0233]**
2000 Pct. Of 25+ Pop. With HS -0.1842[0.0257]**
1940 Pct. Of 25+ Pop. With BA 1.1176[0.1337]**
1940 Pct. Of 25+ Pop. With HS -0.1906[0.0329]**
1850 Pct. Of Pop. Enrolled in College 3.0137 2.4933[0.8351]** [0.8285]**
1850 Pct. Of Pop. Enrolled in HS -0.0499 -0.0129[0.0173]** [0.0194]
1850 Illiteracy Rate 0.061[0.0327]
1850 Pct. Of Pop. Enslaved 0.0372[0.0102]**
Share of labor force in manufacturing, 1950 -0.0447[0.0109]**
Constant 1.3835 1.013 1.5937 1.1751 0.9877 0.7564 0.9174[0.0782]** [0.0960]** [0.1045]** [0.0917]** [0.1311]** [0.1411]** [0.0960]**
Observations 258 282 282 281 208 208 281R-squared 0.39 0.15 0.49 0.32 0.2 0.26 0.19
Notes:Standard errors in brackets. * significant at 5%; ** significant at 1%
2000 Gini Coefficient
Table 2Causes of Urban Inequality
Source: The Gini coefficients are calculated using the 5% Integrated Public Use Microdata Series for 1980 and 2000, at usa.ipums.org. Other 1980 variables are from the 1980 Census, and other 2000 variables are from the 2000 Census. All other variables are from Haines, M.R., ICPSR study number 2896, Historical, Demographic, Economic, and Social Data: The United States, 1790-2000.
40
(1) (2)
1980 Gini Coefficient 0.7934 0.6189[0.0734]** [0.0781]**
Ln(1980 Population) 0.0049 0.0037[0.0014]** [0.0014]**
Ln(1980 Median Family Income) 0.0293 0.0359[0.0126]* [0.0143]*
1980 Pct. Of 25+ Pop. With BA 0.1773[0.0371]**
1980 Pct. Of 25+ Pop. With HS -0.1366[0.0246]**
Constant -0.2165 -0.1352[0.1367] [0.1508]
Observations 242 242R-squared 0.42 0.49
Notes:Standard errors in brackets. * significant at 5%; ** significant at 1%
2006 Gini Coefficient
Table 3Changes in Inequality over Time
Source: The Gini coefficient for 2006 is from the American Community Survey, 2006. The Gini coefficient for 1980 is calculated using the 5% Integrated Public Use Microdata Series (IPUMS) for 1980, at usa.ipums.org. Other 1980 variables are from the 1980 Census.
41
(1) (2) (3) (4) (5) (6)
1980 Gini Coefficient 1.0301 -2.0335 -0.264 -1.3893 50.3572 55.3637[0.7147] [0.6512]** [0.3524] [0.3635]** [14.5159]** [17.1951]**
Ln(1980 Population) 0.0329 0.017 0.0255 0.0275 1.1389 0.9193[0.0141]* [0.0112] [0.0069]** [0.0062]** [0.3450]** [0.3282]**
Ln(1980 Median Family Income) -0.2484 -0.6385 -0.1137 -0.4195 0.8419 8.9701[0.1228]* [0.1141]** [0.0606] [0.0637]** [2.6157] [2.9379]**
1980 Pct. Of 25+ Pop. With BA 1.0094 1.2358 -12.388[0.2995]** [0.1671]** [7.1261]
1980 Pct. Of 25+ Pop. With HS 0.8556 -0.0066 -9.5897[0.1957]** [0.1092] [4.9076]
1994 Mean Jan. Temp. 0.0084 -0.0003 0.0316[0.0009]** [0.0005] [0.0269]
Constant 1.8515 6.0646 1.814 5.0625 -36.9898 -109.0205[1.3318] [1.1903]** [0.6567]** [0.6644]** [29.0251] [31.1609]**
Observations 258 258 258 258 120 120R-squared 0.05 0.4 0.04 0.18 0.2 0.29
Notes:Standard errors in brackets. * significant at 5%; ** significant at 1%
Murder Rate per 100,000Population Growth 1980-2000 Income Growth 1980-2000
Table 4Consequences of Urban Inequality
Source: The Gini coefficient for 1980 is calculated using the 5% Integrated Public Use Microdata Series (IPUMS) for 1980, at usa.ipums.org. Other 1980 variables are from the 1980 Census. January temperatures are from the 1994 County and City Data Book from the U.S. Census.
42
Mobile AL
Jefferson AL
Madison ALMaricopa AZ
Pulaski AR
San Diego CASan Mateo CA
Contra Costa CA
Ventura CA
Santa Cruz CA
Marin CA
Sacramento CA
Los Angeles CASan Francisco CA
Santa Clara CA
Solano CA
Alameda CAOrange CA
San Joaquin CA
Denver CO
Jefferson CO
Arapahoe COBoulder CO
Hartford CT
Middlesex CT
Fairfield CT
Tolland CT
New Haven CT
New London CT
New Castle DE
District of ColumbiaDC
St. Lucie FL
Leon FL
Seminole FL
Volusia FL
Pasco FLBrevard FL
Sarasota FLLee FL
Palm Beach FL
Escambia FL
Broward FLDuval FL
Hillsborough FLOrange FL
Pinellas FL
Manatee FL
Douglas GA
Muscogee GA
DeKalb GA
Clarke GA
Cobb GA
Chatham GA
Forsyth GA
Cherokee GA
Clayton GA
Fulton GA
Rockdale GA
Bibb GARichmond GA
Hall GA
Gwinnett GA
Fayette GAHenry GA
Honolulu HI
McHenry ILWill IL
Madison ILRock Island IL
Cook IL
DuPage IL
Lake IL
Kane ILSt. Clair ILWinnebago ILLake INAllen IN
Vanderburgh IN
Johnson IN
St. Joseph IN
Hamilton IN
Marion IN
Elkhart INPorter IN
Floyd IN
Scott IA
Polk IAWyandotte KSJohnson KS
Sedgwick KS
Campbell KY
Boone KY Kenton KY
Jefferson KYFayette KY
Jefferson LA
Orleans LA
East Baton Rouge LALafayette LA
Baltimore MD
Prince George's MD
Baltimore MD
Carroll MDHarford MDCalvert MD
Montgomery MD
Howard MDAnne Arundel MD
Plymouth MA
Hampden MAEssex MAMiddlesex MA
Barnstable MA
Suffolk MA
Norfolk MA
Worcester MABristol MA
Ingham MI
Kent MI
Kalamazoo MIWayne MI
Macomb MIOttawa MI
Washtenaw MI
Muskegon MI
Genesee MI
Oakland MI
Anoka MN
Washington MN
Ramsey MN
Hennepin MN
Dakota MN
Harrison MSJackson MO
Greene MO
St. Louis MOSt. Louis MO
Clay MOSt. Charles MO
Sarpy NE
Douglas NE
Rockingham NHHillsborough NH
Mercer NJ
Somerset NJMorris NJ
Ocean NJ
Monmouth NJCamden NJ
Middlesex NJ
Union NJPassaic NJ
Hudson NJ
Gloucester NJ
Essex NJ
Bergen NJCape May NJ
Burlington NJ
Atlantic NJBernalillo NM
Kings NY
Orange NY
New York NY
Nassau NY
Richmond NY
Bronx NY
Westchester NY
Dutchess NYPutnam NY
Schenectady NY
Erie NY
Suffolk NY Queens NY
Rockland NYMonroe NY
Niagara NY
Onondaga NYGuilford NC
New Hanover NC
Forsyth NC
Gaston NCWake NC
Cabarrus NC
Durham NC
Cumberland NC
Mecklenburg NC
Catawba NCStark OHMontgomery OH
Cuyahoga OHLucas OH
Hamilton OH
Lake OH
Medina OH
Warren OHClermont OH
Summit OHFranklin OH
Lorain OHButler OH
Clark OHGreene OHTrumbull OH
Mahoning OHOklahoma OK
Cleveland OK
Tulsa OK
Washington OR
Multnomah OR
Lebanon PA
Cumberland PA
Montgomery PALuzerne PA
Lancaster PALehigh PA
Beaver PA
Erie PABucks PA
Allegheny PA
Dauphin PA
Chester PA
Berks PA
Westmoreland PA
Lackawanna PA Delaware PA
Philadelphia PA
York PA
Northampton PAWashington RI
Providence RINewport RI
Kent RI
Richland SC
Greenville SC
Charleston SC
Davidson TNSullivan TNHamilton TN
Washington TN
Shelby TN
Knox TNEl Paso TX
Bexar TX
Rockwall TX
Hidalgo TX
Collin TX
Travis TX Dallas TXCameron TX Harris TX
Fort Bend TX
Nueces TX
Galveston TXGregg TX
Denton TX
Tarrant TX
Weber UTDavis UT
Salt Lake UTNewport News VA
Loudoun VAPrince William VA
Henrico VA
Virginia Beach VA
Chesterfield VA
Portsmouth VA
Alexandria VA
Stafford VA
Lynchburg VA
Hampton VA
Roanoke VANorfolk VAArlington VA
Fairfax VA
Chesapeake VA
Roanoke VA
Richmond VA
Clark WAPierce WA
Island WA
King WA
Kitsap WA
Cabell WV
Milwaukee WI
Brown WI
Dane WIWaukesha WIRacine WI
Winnebago WIOzaukee WI
Kenosha WI
.35
.4.4
5.5
.55
.6G
ini C
oeffi
cien
t, 20
06
-2 0 2 4 6Log of Population Density
Figure 1: Relationship Between the Gini Coefficientand Log Population Density, 2006
Source: 2006 American Community Survey
43
Appleton-O
Athens, GA
Brownsvill
Gainesvill
Janesville
McAllen-Ed
New Haven-
New York,
Ocala, FL
San Jose,
Sheboygan,
Trenton, N
Tuscaloosa
Wausau, WI
Wichita Fa
York, PA
.35
.4.4
5.5
.55
Gin
i Coe
ffici
ent 2
006
.3 .35 .4 .45Gini Coefficient 1980
Figure 2: Gini Coefficient in 2006 and Gini Coefficent in 1980
Notes: The line shown is the forty five degree line, not a fitted regression. Only some datapoints are labeled with their MSA names to aid readability. Source: 1980 Gini coefficients are calculated from the 5% Integrated Public Use Microdata Series (IPUMS) for 1980, at usa.ipums.org. 2006 Gini coefficients are from the 2006 American Community Survey. Source: 2006 American Community Survey
44
Abilene, TAkron, OH
Albany, GA
Albany-Sch
Albuquerqu
Alexandria
Allentown-Altoona, P
Amarillo,
Ames, IA
Anchorage,
Anderson,
Anderson,
Ann Arbor,
Anniston-O
Appleton,
Asheville,
Athens-Cla
Atlanta-SaAtlantic C
Auburn-Ope
Augusta-RiAustin-Rou
Bakersfiel
Baltimore-Bangor, ME
Barnstable
Baton Roug
Battle Cre
Bay City,
Beaumont-P
Bellingham
Bend, ORBillings,
Binghamton
Birmingham
Bismarck,
Blacksburg
Bloomingto
BloomingtoBoise City
Boston-CamBoulder, CBowling Gr
Bremerton-
Bridgeport
BrownsvillBrunswick,
Buffalo-NiBurlington
Burlington
Canton-Mas
Cape Coral
Casper, WYCedar Rapi
Champaign-Charleston
Charleston
Charlotte-
Charlottes
Chattanoog
Cheyenne,
Chicago-Na
Chico, CA
Cincinnati
Clarksvill
Cleveland, Cleveland-
Coeur d'Al
College St
Colorado S
Columbia,Columbia,
Columbus,
Columbus,
Columbus,
Corpus Chr
Corvallis,
CumberlandDallas-For
Dalton, GA
Danville,Danville, Davenport-
Dayton, OH
Decatur, ADecatur, I
Deltona-Da
Denver-Aur
Des Moines
Detroit-Wa
Dothan, AL
Dover, DEDubuque, IDuluth, MN
Durham, NC
Eau Claire
El Centro,
Elizabetht
Elkhart-GoElmira, NY
El Paso, T
Erie, PAEugene-SprEvansville
Fairbanks,
Fargo, ND-Farmington
Fayettevil
Fayettevil
Flagstaff,
Flint, MI
Florence,Florence-M
Fond du La
Fort ColliFort Smith Fort Walto
Fort Wayne
Fresno, CA
Gadsden, A
Gainesvill
Gainesvill
Glens Fall
Goldsboro,
Grand Fork
Grand JuncGrand RapiGreat Fall
Greeley, C
Green Bay,
Greensboro
GreenvilleGreenville
Gulfport-BHagerstown
Hanford-CoHarrisburg
HarrisonbuHartford-W
Hattiesbur
Hickory-Le
Hinesville
Holland-Gr
Honolulu,
Hot Spring
Houma-Bayo
Houston-SuHuntington
Huntsville
Idaho Fall
Indianapol
Iowa City,
Ithaca, NY
Jackson, M
Jackson, MJackson, T
Jacksonvil
Jacksonvil
JanesvilleJefferson
Johnson Ci
Johnstown,Jonesboro,
Joplin, MO
Kalamazoo-
Kankakee-B
Kansas CitKennewick-
Killeen-Te
Kingsport-
Kingston,
Knoxville,
Kokomo, INLa Crosse,
Lafayette,
Lafayette,
Lake Charl
Lakeland, Lancaster,Lansing-Ea
Laredo, TXLas Cruces
Las Vegas-
Lawrence,Lawton, OK
Lebanon, P
Lewiston-A
Lexington-
Lima, OHLincoln, N
Little Roc
Logan, UT-
Longview,
Longview,
Los Angele
Louisville
Lubbock, T
Lynchburg,Macon, GAMadera, CA
Madison, W
ManchesterMansfield,
McAllen-Ed
Medford, O
Memphis, T
Merced, CA
Miami-Fort
Michigan C
Midland, T
Milwaukee-
Minneapoli
Missoula,Mobile, AL
Modesto, C
Monroe, LA
Monroe, MI
Montgomery
Morgantown
Morristown Mount Vern
Muncie, IN
Muskegon-N
Myrtle BeaNapa, CA
Naples-Mar
Nashville-
New Haven-
New Orlean
New York-N
Niles-Bent
Norwich-NeOcala, FL
Ocean City
Odessa, TX
Ogden-Clea
Oklahoma C
Olympia, WOmaha-Coun
Orlando-KiOshkosh-Ne
Owensboro,
Oxnard-ThoPalm Bay-M
Panama Cit
Parkersbur
Pascagoula
Pensacola-Peoria, IL
PhiladelphPhoenix-Me
Pine Bluff
Pittsburgh
PittsfieldPocatello,
Portland-S
Portland-V
Port St. L
Poughkeeps
Prescott, Providence
Provo-Orem
Pueblo, CO
Punta GordRacine, WI
Raleigh-Ca
Rapid City
Reading, P
Redding, C
Reno-SparkRichmond,Riverside-
Roanoke, V
Rochester,
Rochester,Rockford,
Rocky MounRome, GA
SacramentoSaginaw-Sa
St. Cloud,
St. George
St. Joseph
St. Louis,
Salem, OR
Salinas, CSalisbury,
Salt Lake
San AngeloSan Antoni San Diego-
Sandusky,
San Franci
San Jose-SSan Luis O
Santa Barb
Santa Cruz
Santa Fe,
Santa Rosa
Sarasota-BSavannah,
Scranton-- Seattle-Ta
Sebastian-
Sheboygan,
Sherman-De
ShreveportSioux City
Sioux Fall
South BendSpartanbur
Spokane, WSpringfielSpringfiel
Springfiel
Springfiel
State Coll
Stockton,
Sumter, SC
Syracuse,
TallahasseTampa-St.
Terre Haut
Texarkana,Toledo, OH
Topeka, KS
Trenton-Ew
Tucson, AZTulsa, OK
Tuscaloosa
Tyler, TX
Utica-Rome
Valdosta,
Vallejo-Fa
Victoria,
Vineland-MVirginia B
Visalia-PoWaco, TX
Warner Rob
WashingtonWaterloo-C
Wausau, WI
Weirton-St
Wenatchee,
Wheeling, Wichita, K
Wichita Fa
Williamspo
Wilmington
Winchester
Winston-Sa
Worcester,Yakima, WA
York-Hanov
YoungstownYuba City,Yuma, AZ
.35
.4.4
5.5
.55
Gin
i Coe
ffici
ent,
2006
10 10.5 11 11.5Log Median Family Income, 2006
Figure 3: Relationship BetweenGini Coefficient and Log Median Family Income, 2006
Source: 2006 American Community Survey
45
Abilene, T
Akron, OH
Albany, GA
Albany-Sch
Albuquerqu
Alexandria
Allentown-Altoona, P
Amarillo,
Anchorage,
Ann Arbor,
Anniston,
Appleton-O
Asheville,
Athens, GA
Atlanta, GAtlantic-C
Augusta-Ai
Austin-San
Bakersfiel
Baltimore,
Baton Roug
Beaumont-P
BellinghamBenton Har Bergen-PasBillings,
Biloxi-GulBinghamton
Birmingham
Bloomingto
Boise City
Boston-Wor
Bremerton,
Brownsvill
Buffalo-Ni
BurlingtonCanton-Mas
Cedar Rapi
Champaign-
Charleston CharlestonCharlotte-
Charlottes
ChattanoogChicago, I
Chico-Para
Cincinnati
Clarksvill
Cleveland-Colorado S
Columbia,
Columbia,
Columbus,
Columbus,
Corpus Chr
Cumberland Dallas, TXDanville,
Davenport-
Dayton-Spr
Daytona Be
Decatur, I
Denver, CODes Moines
Detroit, MDuluth-Sup
Dutchess C
Eau Claire
El Paso, T
Elkhart-GoErie, PA
Eugene-SprEvansville
Fargo-MoorFayettevil
Fayettevil
Flint, MI
Florence,Florence,
Fort Colli
Fort LaudeFort Myers
Fort Smith
Fort Wayne
Fresno, CA
Gadsden, A
Gainesvill
Galveston-
Gary, IN
Glens Fall
Grand Rapi
Greeley, C
Green Bay,
GreensboroGreenvilleHagerstown
Hamilton-MHarrisburg
Hartford,Hickory-MoHonolulu,Houston, T
HuntingtonHuntsville
IndianapolJackson, M
Jackson, MJacksonvil
Jacksonvil
Janesville
Jersey Cit
Johnson Ci
Johnstown,
Joplin, MO
Kalamazoo-Kankakee,
Kansas Cit
Kenosha, W
Killeen-Te
Knoxville,
Kokomo, INLafayette,
Lafayette,
Lake Charl
Lakeland-W
Lancaster,Lansing-Ea
Las Vegas,
Lexington,
Lima, OHLincoln, N
Little Roc
Longview-M
Los Angele
LouisvilleLubbock, T
Lynchburg,
Macon, GA
Madison, WMansfield,
McAllen-Ed
Medford-As
Melbourne-
Memphis, T
Miami, FL
Middlesex-
Milwaukee-Minneapoli
Mobile, AL
Modesto, C
Monmouth-O
Monroe, LA
Montgomery
Muncie, IN
Nashville,
Nassau-Suf
New Haven-
New London
New Orlean
New York,
Newark, NJ
Newburgh,
Norfolk-Vi
Ocala, FL
Odessa-Mid
Oklahoma COlympia, W
Omaha, NE-
Orange Cou
Orlando, F
Parkersbur
Pensacola,
Peoria-Pek
Philadelph
Phoenix-MePittsburghPortland, Portland-V
Providence
Provo-Orem
Pueblo, CO
Racine, WI
Raleigh-Du
Reading, P
Redding, C
Reno, NV
Richland-K
Richmond-P
Riverside-Roanoke, V
Rochester,
Rockford,
Sacramento
Saginaw-Ba
St. Cloud,
St. Joseph St. Louis,Salem, OR
Salinas, C
Salt Lake
San Antoni
San Diego, San Franci
San Jose,
Santa BarbSanta Cruz
Santa Rosa
Sarasota-BSavannah,
Scranton--Seattle-Be
Sharon, PASheboygan,
Shreveport
Sioux City
Sioux Fall
South Bend
Spokane, WSpringfiel
Springfiel Springfiel
State Coll
Steubenvil
Stockton-L
Syracuse,Tacoma, WA
Tallahasse
Tampa-St.
Terre Haut
Texarkana,
Toledo, OHTopeka, KS
Trenton, NTucson, AZTulsa, OK
Tuscaloosa
Tyler, TX
Utica-Rome
Vallejo-FaVentura, C
Vineland-M
Visalia-Tu
Waco, TX
Washington
Waterloo-C
Wausau, WI
West Palm
Wheeling,
Wichita, KWichita Fa
Williamspo
Wilmington
Wilmington
Yakima, WA
York, PAYoungstown
Yuba City,
.3.3
5.4
.45
Gin
i Coe
ffici
ent 1
980
9.4 9.6 9.8 10 10.2 10.4Log Median Family Income, 1980
Figure 4: Gini Coefficient and Log of Median Family Income, 1980
Source: Gini coefficients are calculated using the 5% Integrated Public Use Microdata Series (IPUMS) for 1980, at usa.ipums.org. Median family income is from the 1980 Census.
46
Abilene, T
Akron, OH
Albany, GA
Albany-Sch
Albuquerqu
Alexandria
Allentown-
Altoona, P
Amarillo,
Anchorage,
Ann Arbor,
Anniston,
Appleton-O
Asheville,
Athens, GA
Atlanta, GAtlantic-C
Auburn-Ope
Augusta-Ai Austin-SanBakersfiel
Baltimore,Barnstable
Baton RougBeaumont-P
BellinghamBenton Har
Bergen-Pas
Billings,Biloxi-Gul
Binghamton
BirminghamBloomingto
BloomingtoBoise City
Boston-WorBoulder-Lo
Brazoria,Bremerton,
Brownsvill
Bryan-Coll
Buffalo-Ni
Canton-Mas
Cedar Rapi
Champaign- Charleston
Charlotte-Charlottes Chattanoog Chicago, I
Chico-ParaCincinnati
Clarksvill
Cleveland-
Colorado S
Columbia,
Columbia,
Columbus,
Columbus,Corpus Chr
Dallas, TXDanville,
Davenport- Dayton-SprDaytona BeDecatur, A
Decatur, I Denver, CODes Moines
Detroit, MDothan, AL
Dover, DE
Duluth-SupDutchess C
Eau Claire
El Paso, T
Elkhart-Go
Erie, PA
Eugene-SprEvansvilleFargo-Moor
Fayettevil
Fayettevil
Flagstaff,Flint, MI
Florence,
Fort Colli
Fort LaudeFort Myers
Fort Pierc
Fort Smith
Fort WaltoFort Wayne
Fort Worth
Fresno, CAGadsden, A
Gainesvill
Galveston-
Gary, IN
Glens Fall
Goldsboro,Grand Junc
Grand Rapi
Greeley, C
Green Bay,
Greensboro
Greenville
Greenville
Hagerstown Hamilton-MHarrisburg
Hartford,
Hattiesbur
Hickory-Mo Honolulu,
Houma, LA
Houston, T
HuntsvilleIndianapol
Iowa City,
Jackson, M
Jackson, MJackson, T
Jacksonvil
Jacksonvil
Jamestown,
Janesville
Jersey CitJohnson Ci
Johnstown,Joplin, MO Kalamazoo-
Kankakee,Kansas Cit
Kenosha, W
Killeen-Te
Knoxville,
Kokomo, IN
La Crosse,
Lafayette,
Lafayette,
Lake Charl
Lakeland-W
Lancaster,
Lansing-Ea
Laredo, TXLas Cruces
Las Vegas,
Lexington,
Lima, OH
Lincoln, NLittle Roc
Longview-M
Los Angele
Louisville
Lubbock, T
Lynchburg,
Macon, GA
Madison, WMansfield,
McAllen-Ed
Medford-As
Melbourne-
Memphis, T
Merced, CA
Miami, FL
Middlesex-
Milwaukee-
Minneapoli
Mobile, AL
Modesto, CMonmouth-O
Monroe, LA
Montgomery
Muncie, IN
Myrtle Bea
Naples, FL
Nashville,
Nassau-Suf
New Haven-
New Orlean
New York,
Newark, NJ
Newburgh,Norfolk-Vi
Oakland, C
Ocala, FL
Odessa-Mid
Oklahoma C
Olympia, W
Omaha, NE-
Orange CouOrlando, F
Panama Cit Pensacola,
Peoria-Pek
Philadelph
Phoenix-Me
Pittsburgh
Portland,
Portland-V
Providence
Provo-Orem
Pueblo, COPunta Gord
Racine, WI
Raleigh-Du
Reading, P
Redding, CReno, NV
Richland-KRichmond-P Riverside-Roanoke, V
Rochester,
Rochester,
Rockford,
Rocky MounSacramentoSaginaw-Ba
St. Cloud,
St. JosephSt. Louis,
Salem, OR
Salinas, C
Salt Lake
San AntoniSan Diego,
San Franci
San Jose,
San Luis OSanta Barb
Santa CruzSanta Fe,
Santa Rosa
Sarasota-B
Savannah,
Scranton--Seattle-Be
Sharon, PA
Sheboygan,
Shreveport
Sioux City
Sioux Fall
South Bend
Spokane, WSpringfielSpringfiel
Springfiel
State Coll
Stockton-LSumter, SC
Syracuse,
Tacoma, WA
Tallahasse
Tampa-St.
Terre HautToledo, OH
Topeka, KS
Trenton, NTucson, AZTulsa, OK
Tuscaloosa
Tyler, TX
Utica-Rome
Vallejo-Fa
Ventura, C
Vineland-MVisalia-Tu
Waco, TX
WashingtonWaterloo-C
Wausau, WI
West Palm
Wichita, K
Wichita Fa
WilliamspoWilmington
WilmingtonYakima, WA
Yolo, CA
York, PA
Youngstown
Yuba City,
Yuma, AZ
.35
.4.4
5.5
.55
Gin
i Coe
ffici
ent 2
000
11 12 13 14 15 16Log of Population
Figure 5: Relationship BetweenGini Coefficent and Log Population, 2000
Source: Gini coefficients are calculated using the 5% Integrated Public Use Microdata Series (IPUMS) for 2000, at usa.ipums.org. Population data is from the 2000 Census.
47
Atlanta, G
Baltimore,Birmingham
Boston-Wor
Buffalo-Ni
Charlotte-
Chicago, I
Cincinnati
Cleveland-
Columbus,
Dallas, TX
Denver, CO
Detroit, M
Fort Worth
Hartford,
Houston, T
Indianapol
Kansas Cit
Los Angele
Memphis, T
Miami, FL
Milwaukee-
Minneapoli
New Orlean
New York,Norfolk-Vi
Oakland, C
Oklahoma C
Orange Cou
Philadelph
Phoenix-Me
Pittsburgh
Portland-V
Providence
Riverside-
Rochester,
Sacramento
St. Louis,
Salt Lake
San AntoniSan Diego, San Franci
San Jose,
Seattle-Be
Tampa-St.
Washington
.22
.24
.26
.28
.3.3
2G
ini C
oeffi
cien
t of H
ousi
ng C
onsu
mpt
ion,
200
0
.4 .45 .5 .55Gini Coefficient 2000
Figure 6:Gini Coefficient of Housing Consumption and the Gini Coefficient, 2000
Source: The 2000 Gini coefficient is calculated using the 5% Integrated Public Use Microdata Series (IPUMS) for 2000, at usa.ipums.org. The housing Gini coefficient was calculated using housing values and controls from the American Housing Survey Metropolitan Samples for 1998, 2002, 2003 and 2004.
48
Akron, OHAlbany-Sch
Albuquerqu
Allentown-
Ann Arbor,Atlanta, G
Austin-SanBakersfiel
Baltimore,
Baton RougBergen-Pas
Birmingham
Boston-Wor
Buffalo-Ni
Charleston
Charlotte-
Chicago, ICincinnatiCleveland-
Colorado S
Columbia,Columbus,
Dallas, TX
Dayton-Spr Denver, CO
Detroit, M
El Paso, T Fort Laude
Fort Wayne
Fort Worth
Fresno, CA
Gary, IN
Grand Rapi
GreensboroGreenville
Harrisburg
Hartford,
Honolulu,
Houston, T
Indianapol
Jacksonvil
Jersey Cit
Kansas Cit
Knoxville,
Las Vegas,
Little Roc
Los Angele
Louisville
McAllen-Ed
Memphis, T
Miami, FL
Middlesex-
Milwaukee-
Minneapoli
Mobile, AL
Monmouth-ONashville,
Nassau-Suf
New Haven-
New Orlean
New York,
Newark, NJ
Norfolk-Vi
Oakland, COklahoma C
Omaha, NE-
Orange CouOrlando, F
Philadelph
Phoenix-Me
Pittsburgh
Portland-V
Providence
Raleigh-DuRichmond-PRiverside-Rochester, Sacramento
St. Louis,
Salt Lake
San AntoniSan Diego,
San Franci
San Jose,
Sarasota-B
Scranton--Seattle-BeSpringfielStockton-LSyracuse,
Tacoma, WA
Tampa-St.Toledo, OH
Tucson, AZTulsa, OK
Vallejo-Fa
Ventura, CWashington
West Palm
Wichita, K
WilmingtonYoungstown
.4.4
5.5
.55
Gin
i Coe
ffici
ent o
f HH
Inco
me,
200
0
.35 .4 .45 .5Gini Coefficient for Male Workers, 2000
Figure 7: Gini for Household Incomeand Gini for Male Workers Ages 25-55, 2000
Source: Gini coefficients are calculated using the 5% Integrated Public Use Microdata Series (IPUMS) for 2000, at usa.ipums.org.
49
Akron, OH
Albany-Sch
Albuquerqu
Allentown-
Ann Arbor,
Atlanta, G
Austin-San
Bakersfiel
Baltimore,Baton Roug
Bergen-Pas
BirminghamBoston-Wor
Buffalo-Ni
CharlestonCharlotte-
Chicago, I
Cincinnati
Cleveland-Colorado S
Columbia,
Columbus,
Dallas, TX
Dayton-Spr
Denver, CO
Detroit, M
El Paso, T
Fort Laude
Fort Wayne
Fort Worth
Fresno, CA
Gary, INGrand Rapi
GreensboroGreenville
Harrisburg
Hartford,
Honolulu,
Houston, T
Indianapol
JacksonvilJersey Cit
Kansas Cit
Knoxville, Las Vegas,Little Roc
Los Angele
Louisville
McAllen-Ed
Memphis, T
Miami, FL
Middlesex-
Milwaukee-Minneapoli
Mobile, AL
Monmouth-ONashville,
Nassau-Suf
New Haven-
New Orlean
New York,
Newark, NJ
Norfolk-Vi
Oakland, C
Oklahoma COmaha, NE-
Orange Cou
Orlando, F
PhiladelphPhoenix-Me
Pittsburgh Portland-V
Providence
Raleigh-DuRichmond-P
Riverside-
Rochester,
Sacramento
St. Louis,Salt Lake
San Antoni
San Diego,
San Franci
San Jose,Sarasota-B
Scranton--
Seattle-Be
Springfiel
Stockton-L
Syracuse,
Tacoma, WA
Tampa-St.
Toledo, OH
Tucson, AZ
Tulsa, OK
Vallejo-Fa
Ventura, C
Washington
West Palm
Wichita, K
Wilmington
Youngstown.35
.4.4
5.5
Gin
i Coe
ffici
ent f
or M
ale
Wor
kers
, 200
0
.18 .2 .22 .24 .26Human Capital Only Gini Coefficient, 2000
Figure 8: Gini Coefficientand Human Capital Only Gini Coefficient, 2000
Source: Gini coefficients are calculated using the 5% Integrated Public Use Microdata Series (IPUMS) for 2000, at usa.ipums.org.
50
Akron, OH
Albany-Sch
Albuquerqu
Allentown-
Ann Arbor,
Atlanta, G
Austin-San
Bakersfiel
Baltimore,Baton Roug
Bergen-Pas
BirminghamBoston-Wor
Buffalo-Ni
CharlestonCharlotte-
Chicago, I
Cincinnati
Cleveland-Colorado S
Columbia,
Columbus,
Dallas, TX
Dayton-Spr
Denver, CO
Detroit, M
El Paso, T
Fort Laude
Fort Wayne
Fort Worth
Fresno, CA
Gary, INGrand Rapi
GreensboroGreenville
Harrisburg
Hartford,
Honolulu,
Houston, T
Indianapol
JacksonvilJersey Cit
Kansas Cit
Knoxville, Las Vegas,Little Roc
Los Angele
Louisville
McAllen-Ed
Memphis, T
Miami, FL
Middlesex-
Milwaukee-Minneapoli
Mobile, AL
Monmouth-ONashville,
Nassau-Suf
New Haven-
New Orlean
New York,
Newark, NJ
Norfolk-Vi
Oakland, C
Oklahoma COmaha, NE-
Orange Cou
Orlando, F
PhiladelphPhoenix-Me
PittsburghPortland-V
Providence
Raleigh-DuRichmond-P
Riverside-
Rochester,
Sacramento
St. Louis,Salt Lake
San Antoni
San Diego,
San Franci
San Jose,Sarasota-B
Scranton--
Seattle-Be
Springfiel
Stockton-L
Syracuse,
Tacoma, WA
Tampa-St.
Toledo, OH
Tucson, AZ
Tulsa, OK
Vallejo-Fa
Ventura, C
Washington
West Palm
Wichita, K
Wilmington
Youngstown.35
.4.4
5.5
Gin
i Coe
ffici
ent f
or M
ale
Wor
kers
, 200
0
.24 .26 .28 .3 .32Human Capital Only Gini Coefficient Using Occupation, 2000
Figure 9: Gini Coefficientand Human Capital Only Gini Coeff. Using Occupations, 2000
Source: Gini coefficients are calculated using the 5% Integrated Public Use Microdata Series (IPUMS) for 2000, at usa.ipums.org.
51
Abilene, TAkron, OH
Albany, GA
Albany-SchAlbuquerqu
Alexandria
Allentown-
Altoona, P
Amarillo,
Ann Arbor,
Anniston,
Appleton-OAsheville,
Athens, GAAtlanta, G
Atlantic-C
Auburn-Ope
Augusta-Ai
Austin-San
Bakersfiel
Baltimore,
Bangor, ME
Barnstable
Baton Roug
Beaumont-P
Bellingham
Benton Har
Bergen-Pas
Billings,
Biloxi-Gul
BinghamtonBirminghamBismarck,
Bloomingto
Bloomingto
Boise City
Boston-Wor
Boulder-Lo
Brazoria,
Bremerton,
Brownsvill
Bryan-Coll
Buffalo-Ni
Burlington
Canton-MasCasper, WY
Cedar Rapi
Champaign-
Charleston
Charleston
Charlotte-
Charlottes
Chattanoog
Cheyenne,
Chicago, I
Chico-Para
Cincinnati
Clarksvill
Cleveland-
Colorado S
Columbia,
Columbia,
Columbus,
Columbus,
Corpus Chr
Corvallis,
Cumberland
Dallas, TX
Danville,
Davenport-Dayton-Spr
Daytona BeDecatur, A Decatur, I
Denver, CO
Des Moines
Detroit, M
Dothan, ALDover, DE
Dubuque, IDuluth-Sup
Dutchess C
Eau Claire
El Paso, TElkhart-Go
Elmira, NYEnid, OKErie, PA
Eugene-Spr
Evansville
Fargo-Moor
Fayettevil
Fayettevil
Flagstaff,
Flint, MIFlorence,Florence,
Fort Colli
Fort Laude
Fort MyersFort Pierc
Fort Smith
Fort Walto
Fort Wayne
Fort Worth
Fresno, CA
Gadsden, A
Gainesvill
Galveston-
Gary, INGlens Fall
Goldsboro,
Grand ForkGrand JuncGrand Rapi
Great FallGreeley, CGreen Bay,Greensboro
Greenville
Greenville
Hagerstown
Hamilton-MHarrisburg
Hartford,
Hattiesbur
Hickory-Mo
Honolulu,
Houma, LA
Houston, T
Huntington
Huntsville
Indianapol
Iowa City,
Jackson, M
Jackson, M
Jackson, T
Jacksonvil
JacksonvilJamestown,Janesville
Jersey Cit
Johnson Ci
Johnstown,
Jonesboro,
Joplin, MO
Kalamazoo-
Kankakee,
Kansas Cit
Kenosha, WKilleen-Te
Knoxville,
Kokomo, IN
La Crosse,
Lafayette,
Lafayette,
Lake CharlLakeland-W
Lancaster,
Lansing-Ea
Laredo, TX
Las Cruces
Las Vegas,
Lawrence,
Lawton, OK
Lewiston-A
Lexington,
Lima, OH
Lincoln, N
Little Roc
Longview-M
Los AngeleLouisville
Lubbock, T
Lynchburg,Macon, GA
Madison, W
Mansfield,McAllen-Ed
Medford-AsMelbourne-Memphis, T
Merced, CA
Miami, FL
Middlesex-
Milwaukee-
Minneapoli Missoula,
Mobile, AL
Modesto, C
Monmouth-O
Monroe, LAMontgomery
Muncie, INMyrtle Bea
Naples, FLNashville,
Nassau-SufNew Haven-
New London
New Orlean
New York,Newark, NJ
Newburgh,Norfolk-Vi
Oakland, C
Ocala, FL
Odessa-Mid
Oklahoma C
Olympia, WOmaha, NE-
Orange Cou
Orlando, F
Owensboro,Panama CitParkersbur
Pensacola,Peoria-Pek
PhiladelphPhoenix-Me
Pine Bluff
PittsburghPittsfieldPocatello,
Portland,
Portland-V
Providence
Provo-Orem
Pueblo, COPunta GordRacine, WI
Raleigh-Du
Rapid City
Reading, PRedding, C
Reno, NVRichland-K
Richmond-P
Riverside-
Roanoke, V
Rochester,
Rochester,
Rockford,
Rocky Moun
Sacramento
Saginaw-Ba
St. Cloud,
St. Joseph
St. Louis,
Salem, ORSalinas, C
Salt Lake
San Angelo
San Antoni
San Diego,
San Franci
San Jose,
San Luis OSanta Barb
Santa CruzSanta Fe,
Santa Rosa
Sarasota-BSavannah,
Scranton--
Seattle-Be
Sharon, PASheboygan,Sherman-DeShreveport
Sioux City
Sioux FallSouth Bend
Spokane, W
Springfiel
SpringfielSpringfiel
State Coll
SteubenvilStockton-L
Sumter, SC
Syracuse,
Tacoma, WA
Tallahasse
Tampa-St.
Terre Haut
Texarkana,
Toledo, OH
Topeka, KS
Trenton, N
Tucson, AZ
Tulsa, OKTuscaloosaTyler, TX
Utica-Rome
Vallejo-Fa
Ventura, C
Victoria,
Vineland-MVisalia-Tu
Waco, TX
Washington
Waterloo-C
Wausau, WI
West Palm
Wheeling,
Wichita, K
Wichita Fa
Williamspo
WilmingtonWilmington
Yakima, WA
Yolo, CA
York, PA
YoungstownYuba City,
Yuma, AZ
.1.2
.3.4
.5S
hare
of A
dults
with
Col
lege
Deg
rees
, 200
0
0 .05 .1 .15Share of Adults with College Degrees, 1940
Figure 10: Relationship Between Share of Adults with College Degrees 2000and Share of Adults with College Degrees 1940
Source: Share of adults with college degrees in 2000 is from the 2000 Census. Share of adults with college degrees in 1940 is from Haines, M.R., ICPSR study number 2896, Historical, Demographic, Economic, and Social Data: The United States, 1790-2000.
52
Abilene, T
Akron, OH
Albany, GA
Albany-SchAlbuquerqu
Alexandria
Allentown-
Altoona, P
Amarillo,
Ann Arbor,
Anniston,
Appleton-O
Asheville, Athens, GA
Atlanta, G
Atlantic-CAuburn-Ope
Augusta-Ai
Austin-San
Bakersfiel
Baltimore,
Bangor, ME
Barnstable
Baton RougBeaumont-P
Bellingham
Benton HarBergen-Pas
Billings,
Biloxi-Gul
Binghamton
Birmingham
Bismarck,Bloomingto
Bloomingto
Boise CityBoston-Wor
Boulder-Lo
Brazoria,
Bremerton,
Brownsvill
Bryan-CollBuffalo-Ni
Burlington
Canton-Mas
Casper, WYCedar RapiChampaign-
CharlestonCharlestonCharlotte-
CharlottesChattanoog
Cheyenne,
Chicago, IChico-Para Cincinnati ClarksvillCleveland-
Colorado SColumbia,
Columbia,
Columbus,
Columbus,
Corpus Chr
Corvallis,
CumberlandDallas, TX
Danville,
Davenport-Dayton-SprDaytona Be
Decatur, A
Decatur, I
Denver, CODes Moines
Detroit, M
Dothan, ALDover, DE
Dubuque, IDuluth-Sup
Dutchess C
Eau Claire
El Paso, T
Elkhart-Go
Elmira, NYEnid, OKErie, PA
Eugene-Spr
Evansville
Fargo-Moor
Fayettevil
Fayettevil
Flagstaff,Flint, MI
Florence,Florence,
Fort Colli
Fort LaudeFort MyersFort Pierc
Fort Smith
Fort WaltoFort Wayne
Fort Worth
Fresno, CA
Gadsden, A
Gainesvill
Galveston-Gary, INGlens Fall
Goldsboro,
Grand ForkGrand Junc Grand Rapi
Great Fall
Greeley, C
Green Bay,
GreensboroGreenville
GreenvilleHagerstown
Hamilton-MHarrisburgHartford,
Hattiesbur
Hickory-Mo
Honolulu,
Houma, LA
Houston, T Huntington
HuntsvilleIndianapol
Iowa City,
Jackson, M
Jackson, M
Jackson, T
Jacksonvil Jacksonvil
Jamestown,Janesville
Jersey CitJohnson Ci
Johnstown,Jonesboro,
Joplin, MO
Kalamazoo-
Kankakee,
Kansas CitKenosha, WKilleen-Te
Knoxville,
Kokomo, IN
La Crosse,
Lafayette,
Lafayette,
Lake CharlLakeland-W
Lancaster,
Lansing-Ea
Laredo, TX
Las Cruces
Las Vegas,
Lawrence,
Lawton, OK
Lewiston-ALexington,
Lima, OH
Lincoln, N
Little Roc
Longview-M
Los Angele
Louisville
Lubbock, T
Lynchburg,
Macon, GA
Madison, W
Mansfield,
McAllen-Ed
Medford-AsMelbourne-
Memphis, T
Merced, CA
Miami, FL
Middlesex-Milwaukee-
MinneapoliMissoula,
Mobile, AL
Modesto, C
Monmouth-O
Monroe, LAMontgomeryMuncie, IN Myrtle BeaNaples, FLNashville,
Nassau-SufNew Haven-
New London
New OrleanNew York,
Newark, NJ Newburgh,Norfolk-ViOakland, C
Ocala, FL
Odessa-Mid
Oklahoma C
Olympia, WOmaha, NE-
Orange Cou
Orlando, FOwensboro,Panama Cit
ParkersburPensacola,Peoria-Pek
PhiladelphPhoenix-Me
Pine Bluff
PittsburghPittsfieldPocatello,
Portland,
Portland-V
Providence
Provo-Orem
Pueblo, COPunta GordRacine, WIRaleigh-Du
Rapid City
Reading, P
Redding, CReno, NV
Richland-KRichmond-P
Riverside-
Roanoke, V
Rochester,
Rochester,Rockford,
Rocky Moun
SacramentoSaginaw-Ba
St. Cloud,
St. Joseph St. Louis,Salem, OR
Salinas, C
Salt Lake
San AngeloSan Antoni
San Diego,San FranciSan Jose,
San Luis O
Santa Barb
Santa CruzSanta Fe,Santa RosaSarasota-B
Savannah,Scranton--
Seattle-Be
Sharon, PASheboygan,
Sherman-DeShreveport
Sioux City
Sioux Fall
South Bend
Spokane, W Springfiel
SpringfielSpringfiel
State Coll
Steubenvil
Stockton-L
Sumter, SC
Syracuse,
Tacoma, WA Tallahasse
Tampa-St. Terre Haut
Texarkana,
Toledo, OH
Topeka, KS
Trenton, NTucson, AZTulsa, OK
TuscaloosaTyler, TX Utica-Rome
Vallejo-FaVentura, C
Victoria,
Vineland-M
Visalia-Tu
Waco, TX
WashingtonWaterloo-CWausau, WIWest Palm
Wheeling,
Wichita, K
Wichita Fa
Williamspo
WilmingtonWilmington
Yakima, WA
Yolo, CA York, PAYoungstown
Yuba City,
Yuma, AZ
.1.2
.3.4
.5S
hare
of A
dult
HS
Dro
pout
s, 2
000
.5 .6 .7 .8 .9Share of Adult HS Dropouts, 1940
Figure 11: Relationship Between Share of Adult HS Dropouts, 2000and Share of Adult HS Dropouts, 1940
Source: Share of adult high school dropouts in 2000 is from the 2000 Census, and share of adult high school dropouts in 1940 is from Haines, M.R., ICPSR study number 2896. Historical, Demographic, Economic, and Social Data: The United States, 1790-2000.
53
Abilene, T
Akron, OH
Albany, GA
Albany-SchAlbuquerqu
Alexandria
Allentown-
Altoona, P
Amarillo,
Anchorage,Ann Arbor,
Anniston,
Appleton-O
Asheville,Athens, GA
Atlanta, G
Atlantic-CAuburn-OpeAugusta-Ai
Austin-San
Bakersfiel
Baltimore,
Bangor, ME
Barnstable
Baton RougBeaumont-P
Bellingham
Benton Har Bergen-Pas
Billings,
Biloxi-Gul
Binghamton
Birmingham
Bismarck,BloomingtoBloomingto
Boise CityBoston-Wor
Boulder-Lo
Brazoria,
Bremerton,
Brownsvill
Bryan-CollBuffalo-Ni
Burlington
Canton-Mas
Casper, WYCedar RapiChampaign-
CharlestonCharlestonCharlotte-
CharlottesChattanoog
Cheyenne,
Chicago, IChico-ParaCincinnatiClarksvillCleveland-
Colorado SColumbia,
Columbia,
Columbus,
Columbus,
Corpus Chr
Corvallis,
Cumberland Dallas, TX
Danville,
Davenport-Dayton-SprDaytona Be
Decatur, A
Decatur, I
Denver, CODes Moines
Detroit, M
Dothan, ALDover, DE
Dubuque, IDuluth-Sup
Dutchess C
Eau Claire
El Paso, T
Elkhart-Go
Elmira, NYEnid, OKErie, PA
Eugene-Spr
Evansville
Fargo-Moor
Fayettevil
Fayettevil
Flagstaff,Flint, MI
Florence,Florence,
Fort Colli
Fort LaudeFort MyersFort Pierc
Fort Smith
Fort WaltoFort Wayne
Fort Worth
Fresno, CA
Gadsden, A
Gainesvill
Galveston-Gary, INGlens Fall
Goldsboro,
Grand ForkGrand JuncGrand Rapi
Great Fall
Greeley, C
Green Bay,
GreensboroGreenville
GreenvilleHagerstown
Hamilton-MHarrisburgHartford,
Hattiesbur
Hickory-Mo
Honolulu,
Houma, LA
Houston, THuntington
HuntsvilleIndianapol
Iowa City,
Jackson, M
Jackson, M
Jackson, T
JacksonvilJacksonvil
Jamestown,Janesville
Jersey CitJohnson Ci
Johnstown,Jonesboro,Joplin, MO
Kalamazoo-
Kankakee,
Kansas CitKenosha, WKilleen-Te
Knoxville,
Kokomo, IN
La Crosse,
Lafayette,
Lafayette,
Lake CharlLakeland-W
Lancaster,
Lansing-Ea
Laredo, TX
Las Cruces
Las Vegas,
Lawrence,
Lawton, OK
Lewiston-ALexington,
Lima, OH
Lincoln, N
Little Roc
Longview-M
Los Angele
Louisville
Lubbock, T
Lynchburg,
Macon, GA
Madison, W
Mansfield,
McAllen-Ed
Medford-AsMelbourne-
Memphis, T
Merced, CA
Miami, FL
Middlesex-Milwaukee-
MinneapoliMissoula,
Mobile, AL
Modesto, C
Monmouth-O
Monroe, LAMontgomeryMuncie, INMyrtle Bea Naples, FLNashville,
Nassau-SufNew Haven-
New London
New OrleanNew York,
Newark, NJNewburgh,Norfolk-Vi Oakland, C
Ocala, FL
Odessa-Mid
Oklahoma C
Olympia, WOmaha, NE-
Orange Cou
Orlando, FOwensboro,Panama CitParkersburPensacola,Peoria-Pek
Philadelph Phoenix-Me
Pine Bluff
PittsburghPittsfieldPocatello,
Portland,
Portland-V
Providence
Provo-Orem
Pueblo, COPunta GordRacine, WIRaleigh-Du
Rapid City
Reading, P
Redding, CReno, NV
Richland-KRichmond-P
Riverside-
Roanoke, V
Rochester,
Rochester,Rockford,
Rocky Moun
SacramentoSaginaw-Ba
St. Cloud,
St. JosephSt. Louis,Salem, OR
Salinas, C
Salt Lake
San Angelo San Antoni
San Diego,San FranciSan Jose,San Luis O
Santa Barb
Santa CruzSanta Fe,Santa RosaSarasota-B
Savannah,Scranton--
Seattle-Be
Sharon, PASheboygan,
Sherman-DeShreveport
Sioux City
Sioux Fall
South Bend
Spokane, WSpringfiel
SpringfielSpringfiel
State Coll
Steubenvil
Stockton-L
Sumter, SC
Syracuse,
Tacoma, WATallahasse
Tampa-St.Terre Haut
Texarkana,
Toledo, OH
Topeka, KS
Trenton, NTucson, AZTulsa, OK
TuscaloosaTyler, TXUtica-Rome
Vallejo-FaVentura, C
Victoria,
Vineland-M
Visalia-Tu
Waco, TX
WashingtonWaterloo-CWausau, WIWest Palm
Wheeling,
Wichita, K
Wichita Fa
Williamspo
WilmingtonWilmington
Yakima, WA
Yolo, CAYork, PAYoungstown
Yuba City,
Yuma, AZ
.1.2
.3.4
.5S
hare
of A
dult
HS
Dro
pout
s, 2
000
0 .2 .4 .6 .8 1Share Hispanic, 2000
Figure 12: Relationship Between Share of Adult HS Dropouts, 2000and Share of Hispanic Population, 2000
Source: 2000 Census.
54
Abilene, T
Akron, OH
Albany, GA
Albany-Sch
Albuquerqu
AlexandriaAllentown-
Altoona, P
Amarillo,Anchorage,
Ann Arbor,
Anniston, Appleton-O
Asheville,
Athens, GA
Atlanta, G
Atlantic-C
Auburn-Ope
Augusta-Ai
Austin-San
Bakersfiel
Baltimore,
Barnstable
Baton Roug
Beaumont-P
Bellingham
Benton Har
Bergen-Pas
Billings,
Biloxi-Gul
Binghamton
BirminghamBloomingto
Bloomingto
Boise City
Boston-WorBoulder-Lo
Brazoria,
Bremerton,
Brownsvill
Bryan-Coll
Buffalo-Ni
Canton-Mas
Cedar Rapi
Champaign-
CharlestonCharlotte-
Charlottes
Chattanoog
Chicago, I
Chico-ParaCincinnati
Clarksvill
Cleveland-
Colorado S
Columbia,
Columbia,
Columbus,
Columbus,Corpus Chr Dallas, TX
Danville,
Davenport-
Dayton-SprDaytona Be
Decatur, ADecatur, I
Denver, CODes Moines
Detroit, MDothan, AL
Dover, DE
Duluth-Sup
Dutchess C
Eau ClaireEl Paso, T
Elkhart-Go
Erie, PAEugene-Spr
Evansville
Fargo-Moor
FayettevilFayettevil
Flagstaff,
Flint, MIFlorence,
Fort Colli
Fort Laude
Fort MyersFort PiercFort Smith
Fort WaltoFort Wayne
Fort WorthFresno, CA
Gadsden, A
Gainesvill
Galveston-
Gary, INGlens Fall
Goldsboro,
Grand Junc
Grand Rapi
Greeley, C
Green Bay,
Greensboro
Greenville
GreenvilleHagerstown
Hamilton-M
HarrisburgHartford,
Hattiesbur
Hickory-Mo
Honolulu,Houma, LA
Houston, T
Huntsville
Indianapol
Iowa City,
Jackson, M
Jackson, M
Jackson, T
Jacksonvil
Jacksonvil
Jamestown,
Janesville
Jersey Cit
Johnson Ci
Johnstown,Joplin, MO
Kalamazoo-Kankakee,
Kansas Cit
Kenosha, W
Killeen-Te
Knoxville,
Kokomo, IN
La Crosse,
Lafayette,
Lafayette,Lake Charl
Lakeland-WLancaster,
Lansing-Ea
Laredo, TX
Las Cruces
Las Vegas,
Lexington,
Lima, OH
Lincoln, N
Little RocLongview-M
Los Angele
Louisville
Lubbock, T
Lynchburg,
Macon, GA
Madison, W
Mansfield,
McAllen-Ed Medford-As
Melbourne-
Memphis, T
Merced, CA
Miami, FL
Middlesex-
Milwaukee-
Minneapoli
Mobile, AL
Modesto, C
Monmouth-O
Monroe, LAMontgomery
Muncie, IN
Myrtle BeaNaples, FL
Nashville,
Nassau-SufNew Haven-
New Orlean
New York,Newark, NJ
Newburgh,
Norfolk-Vi
Oakland, C
Ocala, FL
Odessa-MidOklahoma COlympia, W
Omaha, NE-Orange Cou
Orlando, FPanama Cit
Pensacola,Peoria-Pek
Philadelph
Phoenix-MePittsburghPortland,
Portland-VProvidence
Provo-Orem
Pueblo, CO
Punta GordRacine, WI
Raleigh-Du
Reading, PRedding, C
Reno, NV
Richland-K
Richmond-P
Riverside-
Roanoke, V
Rochester,
Rochester,
Rockford,Rocky Moun
Sacramento
Saginaw-BaSt. Cloud,St. Joseph
St. Louis,
Salem, ORSalinas, C
Salt LakeSan Antoni San Diego,
San Franci
San Jose,
San Luis OSanta Barb
Santa Cruz
Santa Fe,
Santa Rosa
Sarasota-BSavannah,
Scranton--
Seattle-Be
Sharon, PA
Sheboygan,
ShreveportSioux City
Sioux Fall
South BendSpokane, W
Springfiel
Springfiel
Springfiel
State Coll
Stockton-L
Sumter, SC
Syracuse,
Tacoma, WA
Tallahasse
Tampa-St.
Terre HautToledo, OH
Topeka, KS
Trenton, N
Tucson, AZTulsa, OKTuscaloosa
Tyler, TX
Utica-Rome
Vallejo-Fa
Ventura, C
Vineland-MVisalia-Tu
Waco, TX
Washington
Waterloo-CWausau, WI
West Palm
Wichita, KWichita FaWilliamspo
Wilmington
Wilmington
Yakima, WA
Yolo, CA
York, PA
Youngstown
Yuba City,Yuma, AZ
.2.3
.4.5
.6.7
Sha
re o
f Em
ploy
men
t in
Hig
h C
apita
l Ind
ustri
es, 2
000
.1 .2 .3 .4 .5College Completion Among Population 25 or Above, 2000
Figure 13: Relationship Between Share of Employment in High CapitalIndustries and Share of Adults with College Degrees, 2000
Source: 2000 Census
55
Abilene, T
Akron, OHAlbany, GA
Albany-Sch
AlbuquerquAlexandria
Allentown- Altoona, PAmarillo,
Anchorage,
Ann Arbor,
Anniston,
Appleton-OAsheville,
Athens, GAAtlanta, G
Atlantic-C
Auburn-OpeAugusta-Ai
Austin-San
Bakersfiel
Baltimore,
Barnstable Baton Roug
Beaumont-PBellingham Benton Har
Bergen-Pas
Billings,
Biloxi-GulBinghamton
BirminghamBloomingto
Bloomingto
Boise City
Boston-WorBoulder-Lo
Brazoria,
Bremerton,
Brownsvill
Bryan-Coll
Buffalo-Ni
Canton-Mas
Cedar RapiChampaign-
CharlestonCharlotte-
Charlottes
Chattanoog
Chicago, I
Chico-Para
Cincinnati
Clarksvill
Cleveland-
Colorado SColumbia,
Columbia, Columbus,Columbus, Corpus ChrDallas, TX
Danville,
Davenport-
Dayton-Spr
Daytona BeDecatur, ADecatur, I
Denver, CODes MoinesDetroit, M
Dothan, AL
Dover, DE
Duluth-Sup
Dutchess C
Eau ClaireEl Paso, T
Elkhart-GoErie, PAEugene-Spr
Evansville
Fargo-Moor
Fayettevil
Fayettevil
Flagstaff,Flint, MI
Florence,
Fort Colli Fort Laude
Fort MyersFort Pierc
Fort Smith
Fort Walto
Fort Wayne
Fort Worth
Fresno, CA
Gadsden, A
GainesvillGalveston-
Gary, IN
Glens Fall
Goldsboro,Grand JuncGrand RapiGreeley, C
Green Bay,
Greensboro
Greenville
GreenvilleHagerstown
Hamilton-M
Harrisburg
Hartford,
Hattiesbur
Hickory-Mo
Honolulu,
Houma, LAHouston, T
HuntsvilleIndianapol
Iowa City,
Jackson, MJackson, M
Jackson, T
Jacksonvil
Jacksonvil
Jamestown,Janesville
Jersey Cit
Johnson CiJohnstown,
Joplin, MO
Kalamazoo-
Kankakee,
Kansas Cit
Kenosha, W
Killeen-Te
Knoxville,
Kokomo, IN
La Crosse,
Lafayette,Lafayette,
Lake Charl
Lakeland-WLancaster,
Lansing-Ea
Laredo, TXLas Cruces
Las Vegas,
Lexington,
Lima, OH
Lincoln, N
Little RocLongview-M
Los Angele
LouisvilleLubbock, T
Lynchburg,
Macon, GA
Madison, W
Mansfield,
McAllen-EdMedford-As
Melbourne-Memphis, T
Merced, CA
Miami, FL
Middlesex-
Milwaukee-Minneapoli
Mobile, AL
Modesto, C
Monmouth-O
Monroe, LAMontgomery
Muncie, IN
Myrtle BeaNaples, FL
Nashville,
Nassau-SufNew Haven-
New Orlean
New York,Newark, NJ
Newburgh,Norfolk-ViOakland, C
Ocala, FL
Odessa-MidOklahoma C
Olympia, WOmaha, NE-Orange CouOrlando, F
Panama CitPensacola,
Peoria-Pek
Philadelph
Phoenix-MePittsburgh
Portland,
Portland-VProvidence
Provo-Orem
Pueblo, CO
Punta Gord
Racine, WIRaleigh-Du
Reading, PRedding, C
Reno, NV
Richland-K
Richmond-P
Riverside-
Roanoke, V
Rochester,Rochester,
Rockford,
Rocky Moun
Sacramento
Saginaw-Ba
St. Cloud,St. Joseph
St. Louis,
Salem, ORSalinas, C
Salt LakeSan AntoniSan Diego,
San FranciSan Jose,
San Luis OSanta Barb
Santa CruzSanta Fe,
Santa Rosa
Sarasota-B
Savannah,Scranton--
Seattle-Be
Sharon, PASheboygan,
Shreveport
Sioux City
Sioux FallSouth BendSpokane, W
Springfiel
Springfiel
SpringfielState Coll
Stockton-LSumter, SC
Syracuse,
Tacoma, WA
Tallahasse
Tampa-St.
Terre HautToledo, OH
Topeka, KS
Trenton, N
Tucson, AZTulsa, OK
Tuscaloosa
Tyler, TX
Utica-Rome
Vallejo-FaVentura, C
Vineland-M
Visalia-Tu
Waco, TX
Washington
Waterloo-CWausau, WI
West Palm
Wichita, K
Wichita Fa
Williamspo
Wilmington
Wilmington
Yakima, WA
Yolo, CA
York, PA
YoungstownYuba City,
Yuma, AZ
.2.3
.4.5
.6S
hare
of E
mpl
oym
ent i
n Lo
w C
apita
l Ind
ustri
es, 2
000
.5 .6 .7 .8 .9Share of Adults without a College Degree, 2000
Figure 14: Relationship Between Share of Employment in Low CapitalIndustries and Share of Adults without College Degrees, 2000
Source: 2000 Census
56
Akron, OH
Albany-Sch
AlbuquerquAllentown-
Ann Arbor,
Atlanta, GAustin-San
Bakersfiel
Baltimore,
Baton Roug
Bergen-Pas
Birmingham
Boston-Wor
Buffalo-Ni
Charleston
Charlotte-
Chicago, I
Cincinnati
Cleveland- Colorado S
Columbia,Columbus,
Dallas, TX
Dayton-SprDenver, CO
Detroit, M
El Paso, T
Fort Laude
Fort Wayne
Fort WorthFresno, CA
Gary, IN
Grand Rapi
Greensboro
Greenville
Harrisburg
Hartford,
Honolulu,
Houston, T
Indianapol
Jacksonvil
Jersey CitKansas Cit
Knoxville,
Las Vegas,
Little Roc
Los Angele
Louisville
McAllen-Ed
Memphis, TMiami, FL Middlesex-
Milwaukee-Minneapoli
Mobile, AL
Monmouth-ONashville,
Nassau-Suf
New Haven-
New Orlean
New York,Newark, NJ
Norfolk-Vi
Oakland, C
Oklahoma COmaha, NE-
Orange CouOrlando, FPhiladelph
Phoenix-Me
Pittsburgh
Portland-VProvidence
Raleigh-Du
Richmond-P
Riverside-
Rochester,Sacramento
St. Louis,
Salt Lake
San Antoni San Diego,
San FranciSan Jose,
Sarasota-B
Scranton-- Seattle-Be
Springfiel
Stockton-L
Syracuse,
Tacoma, WA
Tampa-St.
Toledo, OH
Tucson, AZ
Tulsa, OKVallejo-Fa
Ventura, C
WashingtonWest Palm
Wichita, K
Wilmington
Youngstown
.3.4
.5.6
.7Lo
g C
olle
ge W
age
Pre
miu
m, 2
000
.1 .2 .3 .4 .5College Completion Among Population 25 or above, 2000
Figure 15:Returns to College and the Percent of Residents with College Degree, 2000
Source: Data is calculated using the 5% Integrated Public Use Microdata Series (IPUMS) for 2000, at usa.ipums.org.
57
Akron, OH
Albany-Sch
Albuquerqu
Allentown-
Ann Arbor,
Atlanta, G
Austin-San
Bakersfiel
Baltimore,Baton Roug
Bergen-Pas
BirminghamBoston-Wor
Buffalo-Ni
CharlestonCharlotte-
Chicago, I
Cincinnati
Cleveland-Colorado S
Columbia,
Columbus,
Dallas, TX
Dayton-Spr
Denver, CO
Detroit, M
El Paso, T
Fort Laude
Fort Wayne
Fort Worth
Fresno, CA
Gary, INGrand Rapi
GreensboroGreenville
Harrisburg
Hartford,
Honolulu,
Houston, T
Indianapol
JacksonvilJersey Cit
Kansas Cit
Knoxville,Las Vegas,Little Roc
Los Angele
Louisville
McAllen-Ed
Memphis, T
Miami, FL
Middlesex-
Milwaukee-Minneapoli
Mobile, AL
Monmouth-ONashville,
Nassau-Suf
New Haven-
New Orlean
New York,
Newark, NJ
Norfolk-Vi
Oakland, C
Oklahoma COmaha, NE-
Orange Cou
Orlando, F
Philadelph
Phoenix-Me
PittsburghPortland-V
Providence
Raleigh-DuRichmond-P
Riverside-
Rochester,
Sacramento
St. Louis,Salt Lake
San Antoni
San Diego,
San Franci
San Jose,Sarasota-B
Scranton--
Seattle-Be
Springfiel
Stockton-L
Syracuse,
Tacoma, WA
Tampa-St.
Toledo, OH
Tucson, AZ
Tulsa, OK
Vallejo-Fa
Ventura, C
Washington
West Palm
Wichita, K
Wilmington
Youngstown.35
.4.4
5.5
Gin
i Coe
ff. fo
r Mal
e W
orke
rs, 2
000
.3 .4 .5 .6 .7Log College Wage Premium, 2000
Figure 16: Gini Coefficient and Returns to College, 2000
Source: Data is calculated using the 5% Integrated Public Use Microdata Series (IPUMS) for 2000, at usa.ipums.org.
58
Akron, OH
Albany-Sch
Albuquerqu
Allentown-
Ann Arbor,
Atlanta, G
Austin-San
Bakersfiel
Baltimore,Baton Roug
Bergen-Pas
BirminghamBoston-Wor
Buffalo-Ni
CharlestonCharlotte-
Chicago, I
Cincinnati
Cleveland-Colorado S
Columbia,
Columbus,
Dallas, TX
Dayton-Spr
Denver, CO
Detroit, M
El Paso, T
Fort Laude
Fort Wayne
Fort Worth
Fresno, CA
Gary, INGrand Rapi
GreensboroGreenville
Harrisburg
Hartford,
Honolulu,
Houston, T
Indianapol
JacksonvilJersey Cit
Kansas Cit
Knoxville,Las Vegas,Little Roc
Los Angele
Louisville
McAllen-Ed
Memphis, T
Miami, FL
Middlesex-
Milwaukee-MinneapoliMobile, AL
Monmouth-ONashville,
Nassau-Suf
New Haven-
New Orlean
New York,
Newark, NJ
Norfolk-Vi
Oakland, C
Oklahoma COmaha, NE-
Orange Cou
Orlando, F
Philadelph
Phoenix-Me
PittsburghPortland-V
Providence
Raleigh-DuRichmond-P
Riverside-
Rochester,
Sacramento
St. Louis,Salt Lake
San Antoni
San Diego,
San Franci
San Jose,Sarasota-B
Scranton--
Seattle-Be
Springfiel
Stockton-L
Syracuse,
Tacoma, WA
Tampa-St.
Toledo, OH
Tucson, AZ
Tulsa, OK
Vallejo-Fa
Ventura, C
Washington
West Palm
Wichita, K
Wilmington
Youngstown.35
.4.4
5.5
Gin
i Coe
ff. fo
r Mal
e W
orke
rs, 2
000
.15 .2 .25 .3 .35Gini Coeff. Holding Skill Constant, 2000
Figure 17: Gini Coefficient and the Gini Coeff. Holding Skills Constant, 2000
Source: Gini coefficients are calculated using the 5% Integrated Public Use Microdata Series (IPUMS) for 2000, at usa.ipums.org.
59
Akron, OH
Albany-Sch
AlbuquerquAllentown-
Ann Arbor,
Atlanta, GAustin-San
Bakersfiel
Baltimore,
Baton Roug
Bergen-Pas
Birmingham
Boston-Wor
Buffalo-Ni
Charleston
Charlotte-
Chicago, I
Cincinnati
Cleveland-Colorado S
Columbia,Columbus,
Dallas, TX
Dayton-SprDenver, CO
Detroit, M
El Paso, T
Fort Laude
Fort Wayne
Fort Worth
Fresno, CA
Gary, IN
Grand Rapi
Greensboro
Greenville
Harrisburg
Hartford,
Honolulu,
Houston, T
Indianapol
Jacksonvil
Jersey CitKansas Cit
Knoxville,
Las Vegas,
Little Roc
Los Angele
Louisville
McAllen-Ed
Memphis, TMiami, FL Middlesex-
Milwaukee-Minneapoli
Mobile, AL
Monmouth-ONashville,
Nassau-Suf
New Haven-
New Orlean
New York,Newark, NJ
Norfolk-Vi
Oakland, C
Oklahoma COmaha, NE-
Orange CouOrlando, F Philadelph
Phoenix-Me
Pittsburgh
Portland-VProvidence
Raleigh-Du
Richmond-P
Riverside-
Rochester,Sacramento
St. Louis,
Salt Lake
San AntoniSan Diego,
San FranciSan Jose,
Sarasota-B
Scranton-- Seattle-Be
Springfiel
Stockton-L
Syracuse,
Tacoma, WA
Tampa-St.
Toledo, OH
Tucson, AZ
Tulsa, OKVallejo-Fa
Ventura, C
WashingtonWest Palm
Wichita, K
Wilmington
Youngstown
.3.4
.5.6
.7Lo
g C
olle
ge W
age
Pre
miu
m, 2
000
.01 .02 .03 .04 .05Share of Employment in Finance, 2000
Figure 18: Returns to Schooling and Share of Workers in Finance, 2000
Source: Gini coefficients are calculated using the 5% Integrated Public Use Microdata Series (IPUMS) for 2000, at usa.ipums.org.
60
Akron, OH
Albany-Sch
AlbuquerquAllentown-
Ann Arbor,
Atlanta, GAustin-San
Bakersfiel
Baltimore,
Baton Roug
Bergen-Pas
Birmingham
Boston-Wor
Buffalo-Ni
Charleston
Charlotte-
Chicago, I
Cincinnati
Cleveland- Colorado S
Columbia,Columbus,
Dallas, TX
Dayton-SprDenver, CO
Detroit, M
El Paso, T
Fort Laude
Fort Wayne
Fort Worth
Fresno, CA
Gary, IN
Grand Rapi
Greensboro
Greenville
Harrisburg
Hartford,
Honolulu,
Houston, T
Indianapol
Jacksonvil
Jersey CitKansas Cit
Knoxville,
Las Vegas,
Little Roc
Los Angele
Louisville
McAllen-Ed
Memphis, TMiami, FL Middlesex-
Milwaukee-Minneapoli
Mobile, AL
Monmouth-ONashville,
Nassau-Suf
New Haven-
New Orlean
New York,Newark, NJ
Norfolk-Vi
Oakland, C
Oklahoma COmaha, NE-
Orange CouOrlando, FPhiladelph
Phoenix-Me
Pittsburgh
Portland-VProvidence
Raleigh-Du
Richmond-P
Riverside-
Rochester,Sacramento
St. Louis,
Salt Lake
San AntoniSan Diego,
San Franci
San Jose,
Sarasota-B
Scranton-- Seattle-Be
Springfiel
Stockton-L
Syracuse,
Tacoma, WA
Tampa-St.
Toledo, OH
Tucson, AZ
Tulsa, OKVallejo-Fa
Ventura, C
WashingtonWest Palm
Wichita, K
Wilmington
Youngstown
.3.4
.5.6
.7Lo
g C
olle
ge W
age
Pre
miu
m, 2
000
0 .02 .04 .06 .08Share of Employment in Computers, 2000
Figure 19: Returns to Schooling and Share of Workers in Computers, 2000
Source: Data is calculated using the 5% Integrated Public Use Microdata Series (IPUMS) for 2000, at usa.ipums.org.
61
Abilene, T
Akron, OHAlbany, GA
Albany-Sch
Albuquerqu
Alexandria
Altoona, PAmarillo,
Anchorage,
Ann Arbor,
Asheville,
Auburn-Ope
Bakersfiel
Baton Roug
Beaumont-P
BellinghamBillings,Binghamton
Bloomingto
Buffalo-Ni
Cedar Rapi
Charleston
Charlottes
Chattanoog
Colorado S
Columbia,
Columbia,
Columbus,
Columbus,
Corpus Chr
Danville,
Decatur, A
Dothan, AL
Dover, DE
Eau Claire
El Paso, TElkhart-GoErie, PA
Eugene-Spr
Fayettevil
Flint, MI
Fort Colli
Fort Smith
Fort Wayne
Fresno, CA
Gadsden, A
Gainesvill
Goldsboro,
Grand Junc
Greeley, CGreen Bay,
Greenville
Honolulu,
Huntsville
Iowa City,
Jackson, M
Jackson, M
Jackson, T
Jacksonvil
Joplin, MO
Kansas Cit
Knoxville,
Kokomo, IN
La Crosse,
Lafayette,
Lafayette,
Lake Charl
Lancaster,
Lansing-Ea
Laredo, TX
Las CrucesLima, OH
Lincoln, N
Little Roc
Lubbock, T
Lynchburg,
Macon, GA
Madison, W
Mansfield,
McAllen-Ed
Merced, CA
Mobile, AL
Modesto, C
Monroe, LA
Montgomery
Muncie, IN
Ocala, FLOklahoma C
Olympia, W
Pittsburgh
Provo-Orem
Pueblo, CO
Punta Gord
Racine, WI
Reading, P
Redding, C
Roanoke, V
Rochester,
Rochester,
Rocky Moun
St. Cloud,
St. Louis,
Salem, ORSalinas, C
San Antoni
Santa Cruz
Santa Fe,
Savannah,
Sheboygan,
Shreveport
Sioux FallSpokane, W
Springfiel
State Coll
Sumter, SC
Syracuse,
Tallahasse
Tampa-St.Toledo, OH
Topeka, KS
Tucson, AZTulsa, OK
Tyler, TX
Utica-Rome
Vineland-MWaco, TX
Waterloo-CWausau, WI
Wichita Fa
WilliamspoWilmington
Yakima, WA
Yuba City,Yuma, AZ
05
1015
Mur
der R
ate
per 1
00,0
00 in
habi
tant
s, 2
000
.35 .4 .45 .5 .55Gini Coefficient, 2000
Figure 20: Murder Rate and the Gini Coefficient, 2000
Source: Gini coefficients are calculated using the 5% Integrated Public Use Microdata Series (IPUMS) for 2000, at usa.ipums.org. Murder rates are from the FBI’s Uniform Crime Reports.
62
Akron, OHAllentown-
Atlanta, G
Austin-San
Baltimore,
Buffalo-Ni
Charleston
Charlotte-
Chicago, ICincinnatiCleveland-
Columbia,
Columbus,
Dallas, TXDayton-Spr
Detroit, MFort Laude
Fort Wayne
Fresno, CA
Gary, IN
Grand Rapi
Houston, T
Indianapol
Jackson, M
Kansas Cit
Knoxville,
Little Roc
Los Angele
Milwaukee-Minneapoli
Nashville,
New Orlean
New York,
Norfolk-Vi
Orlando, FPhiladelph
Phoenix-Me
Pittsburgh
Portland-VRichmond-P
Riverside-
St. Louis,San Diego,
San Franci
Seattle-Be
Syracuse,
Tacoma, WATampa-St.
Tucson, AZ
Washington
West Palm
.8.8
5.9
.95
1S
hare
Sel
f-rep
ortin
g as
Hap
py, 2
000
.4 .45 .5 .55Gini Coefficient 2000
Figure 21: Happiness and the Gini Coefficient, 2000
Source: Gini coefficients are calculated using the 5% Integrated Public Use Microdata Series (IPUMS) for 2000, at usa.ipums.org. Average level of happiness is calculated using data from the General Social Survey.